NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
NIH Public Access
Author Manuscript
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
Published in final edited form as:
Protein Expr Purif. 2005 May ; 41(1): 207–234.
Protein Production by Auto-Induction in High-Density Shaking
Cultures
F. William Studier
Biology Department, Brookhaven National Laboratory, Upton, NY 11973
Abstract
Inducible expression systems in which T7 RNA polymerase transcribes coding sequences cloned
under control of a T7lac promoter efficiently produce a wide variety of proteins in Escherichia
coli. Investigation of factors that affect stability, growth and induction of T7 expression strains in
shaking vessels led to the recognition that sporadic, unintended induction of expression in complex
media, previously reported by others, is almost certainly caused by small amounts of lactose. Glucose
prevents induction by lactose by well-studied mechanisms. Amino acids also inhibit induction by
lactose during log-phase growth, and high rates of aeration inhibit induction at low lactose
concentrations. These observations, and metabolic balancing of pH, allowed development of reliable
non-inducing and auto-inducing media in which batch cultures grow to high densities. Expression
strains grown to saturation in non-inducing media retain plasmids and remain fully viable for weeks
in the refrigerator, making it easy to prepare many freezer stocks in parallel and use working stocks
for an extended period. Auto-induction allows efficient screening of many clones in parallel for
expression and solubility, as cultures have only to be inoculated and grown to saturation, and yields
of target protein are typically several-fold higher than obtained by conventional IPTG induction.
Auto-inducing media have been developed for labeling proteins with selenomethionine, 15N or 13C,
and for production of target proteins by arabinose induction of T7 RNA polymerase from the pBAD
promoter in BL21-AI. Selenomethionine labeling was equally efficient in the commonly used
methionine auxotroph B834(DE3) (found to be metE) or the prototroph BL21(DE3).
Keywords
auto-induction; T7 expression system; lactose; pBAD promoter; arabinose; protein production; highdensity
batch cultures; metabolic control of pH; selemomethionine labeling; isotopic labeling
Background and Introduction
DNA sequencing projects have provided coding sequences for hundreds of thousands of
proteins from organisms across the evolutionary spectrum. Recombinant DNA technology
makes it possible to clone these coding sequences into expression vectors that can direct the
production of the corresponding proteins in suitable host cells. An inducible T7 expression
system is highly effective and widely used to produce RNAs and proteins from cloned coding
sequences in the bacterium Escherichia coli [1, 2]. The coding sequence for T7 RNA
polymerase is present in the chromosome under control of the inducible lacUV5 promoter in
hosts such as BL21(DE3). The coding sequence for the desired protein (referred to as the target
protein) is placed in a plasmid under control of a T7 promoter, that is, a promoter recognized
specifically by T7 RNA polymerase. In the absence of induction of the lacUV5 promoter, little
Correspondence to: F. William Studier.
F. William Studier Biology Department, Bldg 463 Brookhaven National Laboratory PO Box 5000 Upton, NY 11973-5000
studier@bnl.gov Telephone: 631-344-3390 Fax: 631-344-3407
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 2 of 41
T7 RNA polymerase or target protein should be present and the cells should grow well.
However, upon addition of an inducer, typically isopropyl-β-D-thiogalactoside (IPTG)1, T7
RNA polymerase will be made and will transcribe almost any DNA controlled by the T7
promoter. T7 RNA polymerase is so specific, active and processive that the amount of target
RNA produced can be comparable to the amount of ribosomal RNA in a cell. If the target RNA
contains a coding sequence with appropriate translation initiation signals (such as the sequence
upstream of the start codon for the T7 major capsid protein), most protein synthesis will be
directed toward target protein, which usually accumulates to become a substantial fraction of
total cell protein.
A problem in using inducible T7 expression systems is that T7 RNA polymerase is so active
that a small basal level can lead to substantial expression of target protein even in the absence
of added inducer. If the target protein is sufficiently toxic to the host cell, establishment of the
target plasmid in the expression host may be difficult or impossible, or the expression strain
may be unstable or accumulate mutations [3-6]. An effective means to reduce basal expression
is to place the lac operator sequence (the binding site for lac repressor) just downstream of the
start site of a T7 promoter, creating a T7lac promoter [2, 4]. Lac repressor bound at the operator
sequence interferes with establishment of an elongation complex by T7 RNA polymerase at a
T7lac promoter and substantially reduces the level of target mRNA produced [4, 7, 8]. If
sufficient lac repressor is present to saturate all of its binding sites in the cell, the basal level
of target protein in uninduced cells is substantially reduced, but induction unblocks both the
lacUV5 and T7lac promoters and leads to the typical high levels of expression. Thus, the
T7lac promoter increases the convenience and applicability of the T7 system for expressing a
wide range of proteins.
Structural genomics is an area where multi-milligram amounts of many widely different
proteins are sought for determination of protein structures by X-ray crystallography or nuclear
magnetic resonance (NMR) [9]. Not all target proteins will be well expressed and soluble, so
it is desirable to screen in parallel many small cultures expressing different target proteins to
identify those useful for scaling up. A significant difficulty in large-scale screening is to obtain
all of the cultures in a comparable state of growth, so that they can be induced simultaneously.
Differences in lag time or growth rate typically generate a situation where different cultures
will be ready for induction at different times. Even if cultures were grown in a multi-well plate
and densities could be read simultaneously in a plate reader, considerable effort would be
required to follow growth and add inducer to each culture at the proper time. If all of the cultures
were collected at once, choosing a collection time when all had been induced to optimal levels
and none had suffered overgrowth by cells incapable of expressing target protein might be
difficult or impossible.
One strategy for obtaining fairly uniform induction is to incubate a plate until all of the cultures
have grown to saturation, add fresh medium, grow for an appropriate time, and add inducer to
all wells at the same time. If all cultures in a plate saturate at comparable density and grow
after dilution with similar enough kinetics, the culture-to-culture variation in density at the time
of induction might be low enough that most cultures will be optimally induced. However, in
a test of this strategy, I encountered the unintended induction described by Grossman et al.
[6], who found that cultures growing in certain complex media induce substantial amounts of
target protein upon approach to saturation, in the absence of added inducer. Induction at
saturation would stress cells to different extents, depending on the levels of induction and
relative toxicity of target proteins to the host cells, making a strategy of saturation followed by
dilution unworkable in media that have such inducing activity. Grossman et al. [6] concluded
1Abbreviations used: IPTG, isopropyl-β-D-thiogalactoside; PDB, Protein Data Bank; SSAT, human spermidine/spermine
acetyltransferase; SeMet, selenomethionine; TRB, terrific broth; PTS, phosphoenolpyruvate:carbohydrate phosphotransferase system
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 3 of 41
that the known inducer lactose was not responsible for unintended induction but that cyclic
AMP is required, and they found that using a host mutant unable to make cyclic AMP improved
plasmid stability and protein production. Consistent with a role for catabolite repression, they
also found that addition of 1% glucose to the complex medium prevented unintended induction.
However, I observed that addition of 1% glucose also caused saturated cultures to become very
acidic, which limits saturation density and again makes it difficult to get uniform growth upon
dilution.
Upon further investigation, I found that media made with N-Z-amine AS from a 100-pound
barrel recently acquired for structural genomics work showed induction at saturation whereas
otherwise identical media made from the previous (almost exhausted) barrel from the same
supplier did not. Screening different lots of N-Z-amine or other enzymatic digests of casein
for those without the inducing behavior did not seem to be an attractive solution: besides the
obvious inefficiency, such lots might not always be available. To address the problem of
sporadic, unwanted induction, I undertook a systematic analysis of the components of both
complex and defined media and their effects on growth and induction. The goal was to develop
formulations for reliable growth of cultures of T7 expression strains to saturation with little or
no induction and to define conditions suitable for growth and induction of many cultures in
parallel.
Materials and methods
Bacterial strains and plasmids
E. coli strains used for testing growth and expression were primarily BL21(DE3) and B834
(DE3). B834 is a restriction-modification defective, galactose-negative, methionine auxotroph
of E. coli B [10]. BL21 is a Met+ derivative of B834 obtained by P1 transduction [1]. DE3
lysogens contain a derivative of phage lambda that supplies T7 RNA polymerase by
transcription from the lacUV5 promoter in the chromosome [1]. BL21-AI (Invitrogen) is a
derivative of BL21 that supplies T7 RNA polymerase by transcription from the arabinoseinducible
pBAD promoter in the chromosome.
Coding sequences for target proteins were cloned under control of the T7lac promoter and the
upstream translation initiation signals of the T7 major capsid protein [2, 4, 11] by placing the
initiation codon at the position of the NdeI site of pET-13a [12] or pET-24b (Novagen), or the
NcoI site of pREX vectors (equivalent to the NcoI site of pET-11d [2]; to be described
elsewhere), all of which confer resistance to kanamycin. Plasmids containing the T7lac
promoter also contain a copy of the lacI gene to provide enough lac repressor to saturate all of
its binding sites.
A variety of different target proteins were used in developing and testing non-inducing and
auto-inducing media, including a set of about 100 yeast proteins cloned for a structural
genomics project (http://proteome.bnl.gov/targets.html). For convenience, specific yeast
proteins mentioned in the text are referred to by their target numbers: P07 refers to yeast protein
YBL036C, Protein Data Bank (PDB) 1B54, structurally similar to the N-terminal domain of
an amino-acid racemase [13]; P19 refers to yeast protein YBR022W, of unknown function;
P21 refers to the protein specified by yeast gene sup45, a translation release factor; P35 refers
to the protein specified by yeast gene hem13, PDB 1TXN, coproporphyrinogen III oxidase;
and P89 refers to yeast protein YMR087W, PDB 1NJR, proposed from its structure to be an
ADP-ribose-1″-monophosphatase [14]. The coding sequence for human spermidine/spermine
acetyltransferase (SSAT) was amplified by reverse transcriptase and PCR from total RNA from
a human cell line (the kind gift of Paul Freimuth) and cloned in pET-13a. Bacteriophage T7
proteins specified by genes 10A (the well-expressed major capsid protein), 5.3 and 7.7, (highly
toxic proteins of unknown function) [3, 4] were expressed from pREX vectors.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 4 of 41
The expression host for cloned yeast proteins was B834(DE3), in the mistaken belief that a
methionine-requiring host would be better for labeling proteins with selenomethionine (SeMet)
for crystallography (see section on Auto-induction for labeling proteins with SeMet for
crystallography). The RIL plasmid from BL21-Gold(DE3)RIL (Stratagene) increases the
expression of some yeast target proteins by supplying tRNAs for codons used frequently in
yeast but not E. coli. T7 proteins and some other proteins were expressed in BL21(DE3) or
BL21-Gold(DE3)RIL (into which Stratagene introduced the Hte phenotype for high
transformation efficiency and an endA mutation to reduce endonuclease activity). The RIL
plasmid is derived from a pACYC plasmid and confers resistance to chloramphenicol.
Freezer stocks for long-term storage of expression strains are made by adding 0.1 ml of 100%
(w/v) glycerol to 1 ml of culture in log phase or grown to saturation in non-inducing media
such as PG, LSG or MDG (Table 1), mixing well, and placing in a −70°C freezer. Subcultures
for use as working stocks are made by scraping up a small amount of frozen culture with a
sterile plastic pipettor tip without melting the rest of the stock and inoculating into non-inducing
media. After growth to saturation, such working stocks are typically stable for weeks in the
refrigerator.
Growth media
N-Z-amine AS, a soluble enzymatic digest of casein (in 100-pound barrels), and yeast extract
(HY-YEST 444 in a 55-pound barrel) were obtained from Quest International, 5515 Sedge
Blvd., Hoffman Estates, IL 60192, telephone 800-833-8308. For convenience, the designation
N-Z-amine will refer to N-Z-amine AS, which could be substituted for by other enzymatic
digests of casein, such as tryptone, in the media described here. Smaller quantities of enzymatic
digests of casein or yeast extract as well as sugars, salts, amino acids, vitamins and other
components of growth media were obtained from Difco, Sigma, Fisher or other biochemical
and chemical suppliers. Media previously described [1] for growth of E. coli and production
of target proteins with the T7 expression system include ZB (10 g N-Z-amine and 5 g NaCl
per liter), ZYB (previously ZY) (10 g N-Z-amine, 5 g yeast extract and 5 g NaCl per liter), M9
(1 g NH4Cl, 3 g KH2PO4, 6g Na2HPO4, 4 g glucose and 1 ml of 1 M MgSO4 per liter) and
M9ZB, the combination of M9 and ZB. For convenience, concentrations of certain media
components are given in percent (w/v). The previously named ZY medium will here be called
ZYB medium to indicate the presence of 0.5% NaCl, analogous to ZB medium. The name ZY
will be reserved for 1% N-Z-amine, 0.5% yeast extract with no salt added.
The compositions of some of the newly developed media for growing cultures to high density
without induction and for auto-induction are given in Table 1. Media are conveniently
assembled from sterile concentrated stock solutions added to sterile water or ZY just before
use. Standard stock solutions of mixtures include 20xP (1 M Na2HPO4, 1 M KH2PO4, 0.5 M
(NH4)2SO4); 50xL (0.625 M Na2HPO4, 0.625 M KH2PO4, 2.5 M NH4Cl, 0.25 M Na2SO4);
50xM (1.25 M Na2HPO4, 1.25 M KH2PO4, 2.5 M NH4Cl, 0.25 M Na2SO4); 50x5052 (25%
glycerol, 2.5% glucose, 10% α-lactose monohydrate); 100x505 (50% glycerol, 5% glucose).
The term lactose will refer to α-lactose throughout the paper. Stock solutions of individual
compounds include 40% (w/v) glucose; 5% (w/v) aspartic acid neutralized with NaOH; 2.5%
methionine; 1 M disodium succinate; and 1M MgSO4. Heating in a microwave oven is helpful
for dissolving concentrated stock solutions that are slow to dissolve. These stock solutions are
sterilized by autoclaving 15 min and stored at room temperature. The 50xM solution may be
close to saturation or supersaturated; although bottles remained clear for long periods,
occasionally a sample showered crystals, which redissolved readily upon heating in a
microwave oven.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 5 of 41
An amino-acid mixture containing 1% of each of 17 of the 20 natural L-amino acids, lacking
methionine, tyrosine and cysteine, was sterilized by filtration and stored in the refrigerator.
Methionine was omitted for convenience in labeling, tyrosine because it is not soluble enough
to include at this concentration, and cysteine because slow oxidation to the much less soluble
cystine causes precipitate to form. The mixture of 18 amino acids (including methionine but
lacking tyrosine and cysteine) was as effective in promoting growth of BL21(DE3) as a mixture
of all 20 amino acids (an example is given in Table 7). Free amino acids were used to make
the mixture, except for monosodium glutamate, asparagine monohydrate, arginine
monohydrochloride, lysine monohydrochloride and histidine monohydrochloride
monohydrate. The molarity of 0.5% of each amino acid used in the mixture is given in Table
4. When concentrations of amino-acid mixture greater than about 200 μg/ml of each are used,
the amino acids may have to be neutralized with NaOH to keep the pH of the final medium
near neutral.
A stock solution of 0.1 M FeCl3 was dissolved in a 100-fold dilution of concentrated HCl (final
concentration ∼0.12 M HCl). This solution was combined with autoclaved stock solutions of
other metals to make a 1000x trace metal mixture containing 50 mM FeCl3, 20 mM CaCl2, 10
mM each of MnCl2 and ZnSO4, and 2 mM each of CoCl2, CuCl2, NiCl2, Na2MoO4,
Na2SeO3 and H3BO3 in ∼60 mM HCl. These solutions were stored at room temperature. Upon
prolonged storage, small amounts of precipitate formed in the mixture.
Antibiotic stock solutions were kanamycin (25 mg/ml), chloramphenicol (25 mg/ml in ethanol)
and ampicillin (50 mg/ml). Kanamycin was initially used at 25 μg/ml and subsequently at 100
μg/ml (see section High phosphate promotes kanamycin resistance). Chloramphenicol was
used at 25 μg/ml and ampicillin at 50 μg/ml.
The naming convention for media listed in Table 1 and related media is to give a letter
designation to each uniquely different composition of the salts that supply phosphate,
ammonium and sulfate ions (other than MgSO4). P, M and L identify sets of media that supply
100, 50 and 25 mM phosphate, respectively; N and C identify variants used for isotopic labeling
with 15N or 13C. All media contain 2 mM MgSO4 and trace metal mix (although trace metal
mix can be omitted in media containing N-Z-amine and yeast extract). Abbreviations for
complex components, if any, are placed ahead of the letter designation, and abbreviations for
amino acids, glycerol, glucose and lactose are placed after. Thus, Z indicates 1% N-Z-amine,
Y indicates 0.5% yeast extract, and P indicates the salts composition in ZYP medium. The
designation 505 refers to 0.5% glycerol, 0.05% glucose (as in ZYM-505); 5052 refers to 0.5%
glycerol, 0.05% glucose, 0.2% lactose (as in ZYP-5052); and 750501 refers to 0.75% glycerol,
0.05% glucose, 0.01% lactose (in C-750501). G indicates 0.5% glucose, as in PG; D indicates
0.25% aspartate, as in MDG; and A indicates 200 μg/ml of each of 18 different amino acids
(0.36% total amino acids), as in PAG. The S in LSG represents 20 mM succinate and the SM
in PASM is for selenomethionine (SeMet). The names of some media have been shortened
from designations in previously distributed recipes, as indicated in Table 1.
Culture conditions
Cultures were grown in sterile glass vessels in an incubator shaker (New Brunswick G25
series), usually at 300-350 rpm, as indicated on the meter. The incubation temperature was 37•
[unk], unless stated otherwise. Target proteins were expressed at temperatures as low as 18•
[unk]. The standard configuration for growing cultures in parallel was to place 0.5 ml of culture
in 13×100 mm culture tubes with plastic caps. When more than about 0.2 ml of culture was to
be removed for following the time course of growth, pH or induction, 1.5 ml of culture was
grown in 18×150 mm culture tubes or 5-10 ml of culture in 125-ml Erlenmeyer flasks. These
configurations provided sufficient aeration to sustain logarithmic growth to an A600
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 6 of 41
approaching 10 in appropriate media, and expression results seemed to translate well to growth
in 400-500 ml culture volumes in 1.8- or 2.8-liter baffled Fernbach flasks (Bellco), convenient
for producing multi-milligram amounts of proteins in an incubator shaker. Higher rates of
aeration could be obtained with smaller volumes of culture per vessel.
The standard measure of culture growth was optical density at 600 nm (A600) after dilution in
water to concentrations that gave readings below 0.25 in a 1-cm path-length cuvette in a
Beckman DU 640 spectrophotometer. The pH of cultures was measured after 10-fold dilution
in water. Viability and stability of cultures grown under different conditions were tested by
plating on 1% agar plates containing ZB, except as noted. Viable cultures of BL21(DE3)
produced approximately 2 × 109 colonies per ml per A600 over a rather wide range, from log
phase through dense saturated cultures.
Plaque assay for induction of T7 RNA polymerase
To test induction of T7 RNA polymerase in expression hosts in the absence of a target plasmid,
the bacteriophage T7 deletion mutant 4107 was used [1]. This mutant lacks the entire coding
sequence for T7 RNA polymerase and is unable to form a plaque on a lawn of cells unless the
host supplies T7 RNA polymerase. When BL21(DE3) is grown and plated on media that have
no inducing activity, the basal level of T7 RNA polymerase is low enough that only small
plaques develop at low efficiency, and they typically take more than 3 hours to become visible.
In contrast, when BL21(DE3) is induced by including 0.4 mM IPTG in the plate, 4107
efficiently forms the large plaques typical of wild-type T7, which become apparent in less than
2 hours. This 4107 plaque assay was used to test whether T7 polymerase was induced in cultures
of BL21(DE3) grown in different media.
Analysis of proteins on slab gels
Production of target protein was followed by gel electrophoresis of total cell proteins in the
presence of sodium dodecyl sulfate on precast 4-20% polyacrylamide gels (Cambrex). Cells
were lysed in Bugbuster plus Benzonase (Novagen) in 50 mM Tris-Cl, pH 8.0, and containing
egg white lysozyme at 20 μg/ml. Lysozyme improves the release of large proteins into the
soluble fraction but was omitted when it might interfere with identification of proteins of about
the same size in the gel electrophoresis pattern. Benzonase is a DNase that reduces viscosity
that could otherwise interfere with loading samples or cause bands to smear on the gel. Either
a 5x lysis mixture was added directly to an appropriate dilution of culture, or cells were pelleted
by centrifuging 1 min in a micro centrifuge (1.5 ml tubes), the supernatant aspirated, and the
pellet suspended in 1x lysis mixture. The final volume of cell suspension was 40 μl, usually at
a concentration corresponding to a culture density of A600 ∼5, but sometimes half or twice
this concentration. Immediately after mixing, 20 μl of cell suspension was transferred to a
second tube, and both tubes were left for approximately 30 min at room temperature for lysis.
One of the tubes was used as the sample of total cells, to which was added 10 μl of 3x loading
buffer (containing sodium dodecyl sulfate). The other tube was centrifuged 1 min and the
supernatant removed with a pipetter and mixed with another 10 μl of 3x loading buffer to
constitute the soluble fraction. The pellet (insoluble fraction) was suspended in 30 μl of loading
buffer. All three tubes were heated for 1 min in a boiling water bath and 10 μl of each loaded
on the gel for electrophoresis.
Rapid staining of the gel after electrophoresis uses the following protocol. The gel is suspended
in ∼50 ml of 50% ethanol, 10% acetic acid in a covered plastic box, heated almost to boiling
in a microwave oven (with the lid ajar), and then placed on a rocker for at least 5 min at room
temperature, during which the gel shrinks. The liquid is discarded and the gel is suspended in
∼50 ml of 5% ethanol, 7.5% acetic acid plus 200 μl of a 0.25% solution of Coomassie brilliant
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 7 of 41
blue in 95% ethanol. After gentle rocking to disperse the stain, the gel is again heated almost
to boiling in a microwave oven and placed on the rocker. The protein pattern usually becomes
visible within a few minutes and continues to intensify over a few hours. The result can usually
be visualized in less than 30 min but the gel is usually rocked overnight before scanning an
image into the computer. A Kimwipe placed in the solution and rocked for a few minutes can
rapidly take up the slight amount of excess stain in the solvent.
Results
Growth of shaking cultures to high density
Shaking cultures are convenient for growing many cultures in parallel, and rapid growth to
high densities is desirable for maximizing the yield and efficiency of producing target proteins.
Complex media containing enzymatic digests of casein and yeast extract are extensively used
because they support growth of a wide range of E. coli strains with different nutritional
requirements, and cultures typically grow 2-3 times faster than in simple mineral salts media
with glucose as the sole carbon source. However, complex media can vary from lot to lot in
ability to support growth, and some complex media have been found to induce high-level
production of target protein in the T7 expression system upon approach to saturation without
added inducer [6]. To determine what factors might limit growth to high density, and to try to
understand and manage unintended induction, the effects of different components of growth
media on saturation density, growth rate and induction were analyzed.
Results typical of exploratory experiments are shown in Table 2. Cultures of BL21(DE3) grown
overnight in ZB, where 1% N-Z-amine is the sole source of nutrition, saturated at A600 ∼1.2
and pH ∼7.9-8.2. Addition of 0.5% yeast extract (to give ZYB) more than doubled the
saturation density to A600 ∼2.8. Saturation density increased approximately in proportion to
concentration of N-Z-amine up to about 4%, reaching A600 ∼6.9 at 8%. Tripling the
concentration of ZYB almost tripled the saturation density to A600 ∼7.6. Addition of 1%
glucose to ZB, ZYB, 4xZB or 8xZB had little effect on saturation density, apparently because
the acid generated by glucose metabolism overwhelmed the limited buffering capacity of these
media and decreased pH sufficiently to stop growth. Although growth rate was slower in M9
(mineral salts plus 0.4% glucose), the saturation density of A600 ∼2.5 was comparable to that
in ZYB. Adding ZB to M9 tripled the saturation density to A600 ∼7.5, but increasing the glucose
concentration of M9ZB to 2% overwhelmed the buffering capacity of the 66 mM phosphate
buffer in M9 and stopped growth at a lower density, A600 ∼5.8 and pH ∼4.6.
Inducing activity was also analyzed by the ability of BL21(DE3) grown to saturation to support
plaque formation by 4107, a T7 deletion mutant completely unable to form plaques in the
absence of T7 RNA polymerase supplied by the host. Media made with N-Z-amine from our
old barrel (Old ZB in Table 2) had little if any inducing activity. Media made with N-Z-amine
from the new barrel (from which all media were made unless specified otherwise) had
appreciable inducing activity, and higher concentrations of N-Z-amine had higher inducing
activity, as judged by plaque size and time of appearance. Addition of 1% glucose strongly
suppressed inducing activity, as found previously by Grossman et al. [6], but 0.1% glucose
had little effect, presumably because it was depleted well before saturation. This inducing
activity is discussed further in the sections on Non-inducing media and Auto-induction.
Increasing the concentration of N-Z-amine and/or yeast extract can increase saturation density
but can also increase inducing activity and is expensive relative to determining and supplying
precisely what is needed for growth to high density. Simply adding 1 mM MgSO4 to either ZB
or ZY approximately doubled the saturation density (Table 2). Although excess glucose
prevented induction, cultures could become acidic enough to stop growth. Determining and
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 8 of 41
supplying what is needed for growth to high density in batch cultures and understanding and
managing unintended induction has been an iterative process. The following sections
summarize first the growth media that resulted and then the experiments and rationale that led
to them.
High-phosphate P media—Fully defined and complex P media (Table 1) can support the
growth of BL21(DE3) and other E. coli strains to saturation densities of A600 ∼10 or greater
in reasonably well aerated cultures. In P media, an equi-molar mixture of Na2HPO4 and
KH2PO4 provides buffering against metabolically generated changes in pH in both directions
and is a source of sodium, potassium and phosphate ions. A phosphate concentration of 100
mM was chosen to provide as much buffering capacity as possible without stressing the cells.
Higher phosphate levels can be tolerated but growth begins to slow, presumably because of
the high ionic strength. An adequate supply of nitrogen and sulfur is supplied by 25 mM
NH4SO4. The requirement for magnesium ions is satisfied by 1 mM MgSO4, the concentration
given in recipes previously distributed, but the recipes given in Table 1 call for 2 mM
MgSO4, to provide a larger cushion for growth to very high densities. Trace metals are required
for maximal growth in fully defined media. The combined concentration of glucose, glycerol
and other sugars in the recipes given in Table 1 is low enough that they should be depleted
before cultures become irreversibly acidic, and saturated cultures usually have a pH greater
than 6.0. In fully defined media such as PAG and PA-5052, a mixture of 18 purified amino
acids increases growth rate as well as helping to attain approximately neutral pH at saturation.
The standard 200 μg/ml of each amino-acid supported a smooth growth curve to saturation at
densities of at least A600 ∼10, whereas discontinuities were apparent at concentrations of 100
μg/ml or less, presumably because depletion of one or more amino acids required the induction
of synthesis pathways. The doubling time of BL21(DE3) in log-phase growth at 37°C ranged
from about 60-70 min in minimal media to about 30-35 min in media containing ZY or the
mixture of 18 purified amino acids. The recipes for P media have been widely distributed and
used successfully to grow stable stock cultures of T7 expression strains and to produce target
proteins by auto-induction.
High phosphate promotes kanamycin resistance—Expression vectors that confer
resistance to kanamycin were selected for our structural genomics work, to avoid possible
overgrowth of induced cultures by cells that have lost plasmid. Such overgrowth can occur
when expression vectors confer resistance to ampicillin, because secreted β-lactamase can
degrade all of the ampicillin in the medium [1, 2]. However, I was surprised to find that BL21
(DE3) without any plasmid grew to high density overnight at 37°C in auto-inducing ZYP-5052
medium containing 25 μg/ml of kanamycin, a concentration that efficiently kills them in ZB
or ZYB cultures or plates. The cultures that grew had typical plating efficiencies and remained
sensitive to 25 μg/ml of kanamycin in ZYB plates. Furthermore, BL21(DE3) plated directly
on ZP or ZPG plates containing 25 μg/ml of kanamycin formed smaller but uniform colonies
at normal efficiency, indicating that all cells survived and grew in these media.
Systematic tests revealed that the increased resistance to kanamycin is due to high
concentrations of phosphate combined with amino acids and perhaps other nutrients in rich
media. At a kanamycin concentration of 25 μg/ml, BL21(DE3) did not grow in ZYB, which
has no added phosphate, nor in the minimal PG, which contains 100 mM phosphate, but it grew
quite well in the fully defined PAG, which contains both 100 mM phosphate and purified amino
acids. Growth at 25 μg/ml was also observed in other media that contain relatively high
concentrations of phosphate and amino acids, such as M9ZB (64 mM phosphate) and terrific
broth (89 mM phosphate) [15] (here abbreviated TRB to avoid confusion with tryptone broth
(TB)). In rich media, the higher the concentration of phosphate, the higher the concentration
of kanamycin needed to prevent growth and kill cells: BL21(DE3) failed to grow in M9ZB
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 9 of 41
and TRB at 50 μg/ml and was killed effectively at 100 μg/ml; PAG cultures still became turbid
at a kanamycin concentration of 50 μg/ml, killing was somewhat faster than growth at 100
μg/ml and killing was effective at 200 μg/ml; ZYP-5052 cultures still became turbid at 100
μg/ml, killing was slightly faster than growth at 200μg/ml, and killing was effective at 400
μg/ml. Although many uninduced expression strains are relatively stable even in the absence
of selective antibiotic, having rich media in which BL21(DE3) is more sensitive to kanamycin
seemed preferable to resorting to concentrations as high as 400 μg/ml when selection is needed.
A few attempts to develop an amino-acid mixture that would promote rapid growth without
substantially increasing kanamycin resistance were not successful. Reducing the phosphate
concentration in growth media seemed the most attractive way of increasing the sensitivity to
kanamycin.
Lower phosphate M and L media—As described in the next section, cultures can be grown
to high densities with only minimal buffering of pH by phosphate or other buffers. The M and
L sets of media (Table 1) have phosphate concentrations of 50 mM and 25 mM respectively.
Their salt composition was modified from that used in P media to allow independent variation
of phosphate, sulfate and ammonium ions, which is useful for testing nutritional requirements
and for isotopic labeling. Non-inducing and auto-inducing L media (25 mM phosphate) have
been tested extensively and are satisfactory for most purposes, but the M media (50 mM
phosphate) have smaller variations in pH during growth and are currently used for routine
work. BL21(DE3) is killed about as fast as it divides in ZYM-5052 containing kanamycin at
50 μg/ml and is killed fairly effectively at 100 μg/ml. A kanamycin concentration of 100 μg/
ml was adopted for routine work.
Metabolic control of pH—Cultures growing in media containing glucose (and in which no
other nutrient is limiting) will continue to grow until the glucose becomes depleted or the acid
generated by the metabolism of glucose exceeds the buffering capacity of the medium and
causes the pH to drop to a level that stops growth. As long as sufficient glucose is present in
the growth medium, catabolism of other carbon and energy sources that could balance the acid
generated by metabolism of glucose is prevented by the phosphenolpyruvate:carbohydrate
phosphotransferase system (PTS), acting through catabolite repression and inducer exclusion
[16-20]. In the absence of glucose, glycerol can support growth about as effectively but
suppresses the use of other carbon sources less dramatically than glucose by a mechanism
affecting cyclic AMP production [21]. Excess glycerol can also generate enough acid to stop
growth, but, in contrast to glucose, the presence of glycerol does not suppress the inducing
activity found in complex media.
Another factor with a profound influence on growth is the availability of oxygen. If the culture
becomes dense enough that oxygen consumption exceeds the rate of aeration in the shaking
vessel, oxygen limitation triggers complex regulatory responses that attempt to adjust the
metabolic capacities of the cell to the availability of oxygen and the carbon and energy sources
in the medium [22]. The higher the rate of aeration (or oxygenation) the higher the culture
density attained before oxygen limitation triggers these responses. Acid production from
glucose or glycerol seems to increase as a result of the metabolic changes as oxygen becomes
limiting.
Imbalances in needs for energy and carbon in growth with glucose as carbon and energy source
are typically rectified by excretion of acetate and other compounds into the medium [23-25].
If glucose is depleted before the medium gets too far out of balance, the excreted acetate and
other carbon and energy sources that may be present in the medium can then be metabolized,
which can return the pH of the medium to the neutral or alkaline range. The decrease of pH
upon metabolism of glycerol can also be reversed by metabolism of other carbon and energy
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 10 of 41
sources in the medium. Excursions of external pH outside the neutral range on either the acid
or alkaline side also induce complex regulatory responses [26].
The stringent control of the order of catabolism of different carbon and energy sources in the
growth medium, together with the complex regulatory responses to other environmental
conditions, make it challenging to develop media in which the pH remains in a range that
supports growth to high cell densities in shaking vessels. The 100 mM phosphate in P media
provides enough buffering capacity to allow growth to depletion of 0.5% glucose with a
saturation density greater than A600 ∼5 while maintaining pH above 6.0. However, significant
increases in glucose concentration or decreases in phosphate concentration usually produced
cultures that saturated at low pH and lost viability within hours or days. In an attempt to provide
a stronger buffer against decreasing pH, which would allow the use of higher glucose and
glycerol concentrations or lower phosphate concentrations, organic acids with relatively high
pKa were tested for their ability to buffer the medium and thereby allow growth to higher
density.
Organic acids Succinate was found to be effective in countering the acid generated by glucose
in minimal L medium (which has only 25 mM phosphate). It is apparent from results shown
in Table 3 that, rather than acting simply as a buffer, succinate is metabolized as glucose nears
depletion during growth: cultures reach a higher saturation density and a higher pH than in the
absence of succinate. The growth rate and the changes in pH during growth (not shown) are
consistent with glucose being metabolized first and then succinate, as glucose is depleted.
Approximately 20 mM succinate seems optimal for balancing 0.5% (28 mM) glucose, usually
producing saturated cultures with a pH close to neutral. Substantially higher concentrations of
succinate can cause the pH to rise well beyond 8.0, which can stress the cells and reduce
viability. The presence of succinate does not cause detectable induction of T7 RNA
polymerase, as measured by the 4107 plaque assay and as indicated by the viability and stability
of saturated cultures of strains that express highly toxic target proteins. Cultures that saturate
between pH ∼6 and ∼7.5 are stable for weeks in the refrigerator with little loss of viability or
increase in lag time when growing subcultures. Fumarate, DL-malate and citrate were also able
to balance the acid produced by glucose in much the same way as succinate. Added acetate
was effective to a lesser extent. Maleate provided some buffering against the drop in pH but
was toxic to BL21(DE3) at low pH, at least in some media.
Amino acids N-Z-amine, yeast extract or a mixture of 18 pure amino acids (no Y, C) increase
both growth rate and saturation density of glucose- or glycerol-containing media. Uptake of
amino acids from the medium and incorporation directly into proteins spares the cells from
having to make enzymes for entire metabolic pathways and divert carbon from glucose into
synthesis of proteins rather than production of energy or other metabolites. If the concentration
of amino acids is high enough, at least some of them will remain to be catabolized for carbon
and energy after glucose is depleted, causing pH to rise and potentially balancing acid generated
from glucose. In contrast to N-Z-amine, purified amino acids contributed no inducing activity
when added to defined media.
To determine which amino acids are most effective in balancing pH, each of the 18 pure amino
acids used in the mixture was tested individually at a concentration of 0.25% for ability to
balance the acid generated by 0.5% glucose in L medium (25 mM phosphate) (Table 4). The
most effective single amino acid was aspartate, followed by serine, asparagine, glycine and
glutamate, all of which increased the saturation density 60% to115% and produced a pH >6.2
at saturation (compared with pH ∼4.1 in glucose itself). By comparison, 20 mM succinate
(0.24%) increased saturation density by 90% and produced a pH of 6.8 at saturation, and the
mixture of 18 amino acids (0.27%) increased saturation density 75% and produced a pH ∼5.7.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 11 of 41
Of the other amino acids, only glutamine and proline produced as much as a 25% increase in
saturation density and only alanine produced a pH >4.0 at saturation. Of the amino acids that
failed to balance pH at a concentration of 0.25%, only alanine was effective in balancing pH
when tested at 0.5%. Several amino acids substantially increased the lag or decreased the
growth rate in minimal LG medium, most notably serine, alanine, leucine and valine,
presumably by repressing overlapping metabolic pathways [27, 28]. Addition of 0.01% each
of leucine, isoleucine and valine restored normal growth in the presence of 0.25% serine. (The
slow growth in histidine may reflect a low pH of the medium.)
To determine which amino acids are most effectively utilized as a carbon and energy source
for BL21(DE3), cultures were grown in L medium with amino acids as sole carbon source
(Table 4). A mixture of the 18 amino acids, each at 100 μg/ml (0.18% total amino acids) was
provided to promote some growth and to alleviate possible inhibitory effects of individual
amino acids, which were added at a concentration of 0.5%. The mixture of 18 amino acids by
itself supported growth to A600 ∼1.4 with a final pH ∼6.8. Of the individual amino acids,
proline was the most effective carbon and energy source, supporting growth to A600 ∼9.6 and
pH ∼7.0, comparable to A600 ∼9.0 and pH ∼5.7 supported by 0.5% glycerol. Other amino
acids that substantially increased the saturation density were serine, glutamate, alanine and
aspartate, with smaller increases by threonine and asparagine. Each of these amino acids also
increased the final pH at least somewhat, indicating that they were metabolized. The final pH
∼5.2 of the histidine-containing culture represented a substantial decrease from an initial pH
estimated to be ∼6.0 by reconstitution (versus ∼ 6.6 for the other amino acids), suggesting
that metabolism of histidine decreases the pH of the culture. The remaining individual amino
acids did not significantly affect either A600 or pH, suggesting that they were not significantly
catabolized. A credible test of tryptophan was not done.
Minimum nutritional requirements for growth to high density—Metabolic balancing
of pH made it possible to test the requirement for any nutrient including phosphate to support
the growth of BL21(DE3), without the complication of the culture becoming too acidic or basic
for optimal growth. A series of tests of mineral salts media with glucose or glycerol as primary
carbon source established nutrient concentrations that limit growth to low densities, which
could be extrapolated to determine approximate minimum concentrations needed for growth
to high saturation densities. Table 3 shows results of one series of tests of minimal requirements
for sulfur, nitrogen, phosphate and magnesium in modified LG medium, in the absence or
presence of 25 mM succinate. The cultures were inoculated with a thousand-fold dilution of
BL21(DE3) that had been grown to saturation in PG, and 0.5 ml cultures were grown in 13×100
mm tubes in a shaking incubator for 14-15 hr at 37°C. Conclusions from these and similar
experiments are summarized in the following sections.
Sulfur Carryover of 0.026 mM sulfate in the inoculum supported growth to A600 ∼0.7 with
pH ∼6.7. The need for sulfate saturated at approximately 0.5 mM, in which BL21(DE3) grew
to A600 ∼6.1 at pH ∼6.7. A sulfate concentration of 0.5 mM or greater at near neutral pH was
also enough to produce very stable cultures, as measured by viability after three weeks in the
refrigerator. The 5 mM Na2SO4 in L and M media and 25 mM (NH4)2SO4 in P media should
supply enough sulfur to support growth to very high densities in shake flasks.
Nitrogen Saturation density continued to increase with NH4Cl concentration until at least 50
mM, which supported growth to A600 ∼5.5 at pH ∼7.1. Cultures retained high viability for at
least three weeks in the refrigerator at NH4Cl concentrations of 20 mM or higher and pH near
neutral. In minimal media in which pH was maintained near neutral, 50 mM NH4
+ reproducibly
supported growth to slightly higher density than 25 mM and is therefore the standard
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 12 of 41
concentration used in P, M and L media. However, 25 mM NH4Cl is sufficient for most
purposes, including labeling of proteins with 15N for NMR studies.
Phosphate Carryover of 0.1 mM phosphate in the inoculum supported growth to A600 ∼0.8
at pH ∼6.7. The presence of 1 mM phosphate in the medium supported growth to A600 ∼3.8
at pH ∼6.4 but the need for phosphate did not appear to saturate until 10-15 mM at A600 ∼5.9
and pH ∼8.2. E. coli cells have complex regulatory responses when phosphate becomes
limiting in the medium [29], and alternative uses of internal phosphate may account for the
relatively slow increase in saturation density between 1 mM and 10 mM phosphate. The
buffering capacity of phosphate in the medium did not significantly reduce the pH increase
due to succinate metabolism until 35-50 mM phosphate. The minimum phosphate
concentration in the media given in Table 1 is 25 mM, to try to avoid a phosphate limitation
that would induce response mechanisms. Experiments in which saturation densities were
pushed well above A600 ∼10 have occasionally suggested that even 25 mM phosphate may
become limiting at densities achievable in shaking vessels.
Magnesium No growth of BL21(DE3) was apparent in the absence of magnesium, but,
interestingly, cultures containing only limiting amounts of magnesium grew to much higher
densities (5- to 10-fold) when the growth medium contained succinate than when it did not.
The need for magnesium appeared to saturate at 0.5 mM, which gave A600 ∼6.4 and pH ∼6.2.
However, viability after three weeks in the refrigerator seemed to remain somewhat higher in
cultures grown in 1-2 mM MgSO4 than in those grown at lower concentrations. Magnesium
levels as high as 10 mM (the highest concentration tested) showed no inhibition of growth.
Previously distributed recipes for P medium contain 1 mM magnesium, but 2 mM (as given
in Table 1) may provide a greater margin for growth to very high densities.
Trace metals Fully defined media made from purified components contain contaminating trace
metals in amounts sufficient to support growth to moderate density but not sufficient for growth
to high density with good expression of target proteins by auto-induction. Table 5 summarizes
results from an auto-induction experiment to test the effects of trace metals. In this experiment,
the expression strain saturated in ZYP-5052 at A600 ∼18 with the target protein expressed at
high level. In slightly modified PA-5052 without added trace metals, saturation was at A600
∼4.4 with little expression of target protein. Addition of trace metals about tripled the saturation
density, to A600 ∼13, and allowed high-level expression of target protein. Clearly, a deficiency
of trace metals limited culture growth and auto-induction of target protein in this fully defined
medium.
Individual metal ions were tested at concentrations of 1, 10 and 100 μM for ability to increase
saturation density and for possible toxicity (Table 5). The trace metals were chosen as being
likely to have a functional association with proteins or participate in some biological process.
Iron ions at 10 and 100 μM increased saturation density to A600 ∼13 but 1 μM increased the
density only to A600 ∼7.8. Manganese ions at 1, 10 and 100 μM also increased saturation
density to A600 ∼13, as did cobalt ions at 1 and 10 μM. However, 10 μM cobalt ions caused
a lag of about an hour before attaining normal growth rate, and 100 μM cobalt prevented
growth. Zinc ions appeared to have only a slight stimulatory effect, and nickel, molybdate,
calcium, copper, selenate or borate even less. Selenate did not appear to be toxic at 10 μM but
prevented growth at 100 μM.
Many proteins of unknown function are being produced in structural genomics projects, any
one of which might have an unsuspected metal ligand. Target proteins of 50,000 Da produced
at 100 mg/liter would have a concentration of 2 μM and proteins of 10,000 Da a concentration
of 10 μM. The 1x concentration of metal mix supplies 50 μM iron, 20 μM calcium, 10 μM
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 13 of 41
manganese and zinc, and 2 μM cobalt, copper, nickel, molybdate, selenate and borate, amounts
that are not toxic to growth but could saturate potential binding sites in many target proteins.
Of course, if a target protein is known to have a metal ligand, the appropriate concentration of
that specific metal can be added. Concentrations between about 0.1x and 2x metal mix
supported maximum saturation density, 5x was slightly inhibitory and 10x markedly slowed
growth but the culture still attained high density and a high level of auto-induction.
Iron Concentrations of 0.05x metal mix or lower did not support growth to high density in
defined media and produced only low levels of target protein by auto-induction, primarily due
to a deficiency in iron. In the presence of 0.02x metal mix, an iron concentration of 5 μM was
sufficient for maximum growth and auto-induction in a defined medium without amino acids
but 10 μM was needed in the presence of amino acids. The highest iron concentration tested,
500 μM, showed no evidence of toxicity. In a defined medium containing 100 μM FeCl3,
omission of the metal mix only slightly diminished the maximum density and the level of target
protein produced by auto-induction, so 100 μM FeCl3 may suffice for many purposes if a
suitable metal mix is not available.
In contrast to the results summarized in Table 5, manganese or cobalt, alone or in combination,
did not compensate for a deficiency in iron in subsequent experiments. A difference was that
the media used in the tests reported in Table 5 contained seven different vitamins but subsequent
experiments contained no added vitamins. Whether the presence of vitamins could account for
the difference has not been tested.
Complex media Tests of nutritional requirements for growth of BL21(DE3) to high density
in complex media indicate that media containing only ZY are deficient in magnesium,
phosphate, carbon and energy sources, as well as the ability to buffer pH changes that occur
during growth. The high concentrations of amino acids in ZY are almost guaranteed to provide
sufficient nitrogen and sulfur, but the known variability from lot to lot makes it seem prudent
to add 0.2x metal mix, or at least 10 μM of an iron salt, to ensure that trace metal requirements
are met. The mineral salts components of P, M or L media are included in all formulations of
complex media in Table 1 to ensure that minimal requirements for growth to high density and
auto-induction are met.
Fully defined media have been formulated with well-metabolized amino acids at concentrations
high enough to achieve saturation densities equal to or greater than those obtained in complex
media. However, yeast extract appears to supply something that allows slightly more rapid
initial growth than in those fully defined media. Addition of vitamins, purines and pyrimidines
to the defined media had little effect on growth rate or saturation density. Yeast extract supplies
a variety of metabolites, including fats and complex carbohydrates, any of which might be
responsible for a slightly faster initial growth rate.
Non-inducing media
Besides our new barrel of N-Z-amine, a sample of Bacto tryptone (Difco) also had inducing
activity, suggesting that inducing activity may be fairly common in enzymatic digests of casein.
Addition of excess glucose to complex media that have inducing activity prevents induction
of target protein [6], but cultures eventually become acid enough to stop growth and can lose
viability. At intermediate glucose concentrations, cultures became induced if the pH rose at
saturation, indicating that glucose was depleted, but not if the culture stayed acid, indicating
that glucose remained in the culture. The rate of aeration also had a substantial effect on
saturation density, acidity and induction. It seemed difficult or impossible to formulate complex
media in which cultures reliably grow to saturation without induction and don't become so acid
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 14 of 41
as to reduce viability. Therefore, the non-inducing media given in Table 1 are fully defined,
made with purified components that have no detectable inducing activity.
We currently use MDG medium for routine growth of non-induced cultures of BL21(DE3)
expression strains but have previously used PG and LSG extensively for this purpose. These
media support the growth of BL21(DE3) with a doubling time of approximately an hour. Being
minimal media, they must be appropriately supplemented when growing strains with
nutritional requirements, such as B834(DE3). Overnight cultures typically saturate at A600
∼5-9 and a pH near neutral without detectable induction of target protein. When grown to
saturation in these media, even strains that express highly toxic target proteins remain stable
and viable for weeks in the refrigerator, and subcultures grow with little or no lag. This makes
it convenient to grow both freezer stocks and working cultures overnight to saturation, whereas
previously we tried to collect cultures in log phase to minimize potential instabilities if the
target protein is toxic to the host. The cells that settle out of working cultures stored in the
refrigerator usually disperse readily, but occasionally they have been sticky and more difficult
to disperse. The reason for this occasional stickiness has not been determined but may be
associated with a slightly alkaline pH in the saturated culture.
Agar plates made with fully defined non-inducing media such as MDAG or PAG enabled the
isolation of some BL21(DE3) transformants that were unable to form colonies on the ZYB
plates we usually use for selection. Apparently the inducing activity in ZYB plates caused
enough expression of highly toxic target protein to prevent colony formation, but the lack of
inducing activity in the MDAG or PAG plates allowed colonies to form. MDAG or PAG plates
are rich enough that innocuous clones form colonies on them almost as rapidly as on ZYB
plates.
Auto-induction
Unintended induction is almost certainly due to lactose in the medium—Media
made with N-Z-amine from the old barrel did not have inducing activity. Apparently,
something in the new N-Z-amine was causing induction (rather than something in the old NZ-amine
preventing induction) because increasing the concentration of new N-Z-amine in the
medium also increased the inducing activity, as judged by 4107 plaque size and time of
appearance (Table 2). Grossman et al. [6] had concluded that unintended induction was not
due to the presence of lactose in the medium. However, it seemed reasonable to test whether
addition of lactose to media made with N-Z-amine from the old barrel would produce inducing
behavior similar to that observed in media made from the new barrel. Indeed, the results
summarized in Table 6 show that it does. As expected, no induction of B834(DE3)P35 was
apparent in the absence of added lactose. The smallest concentration of lactose tested in this
set, 0.005% (139 μM), gave a high level of induction of P35 protein, but the culture density,
viability and maintenance of plasmid were all comparable to what was found in the absence
of added lactose. Apparently, P35 protein is not very toxic to the cell. With increasing amounts
of lactose, production of P35 protein remained high and the density of the saturated cultures
decreased somewhat, but the viability decreased substantially, particularly at 0.05% lactose
and higher. At these higher lactose concentrations, most of the surviving cells had lost the
expression plasmid. High levels of induction are known to kill cells that carry a multi-copy
plasmid with a T7 promoter, even if the target protein is innocuous [1, 3].
Other experiments (not shown) found that production of P35 protein was still appreciable with
as little as 0.003% (83 μM) lactose, and detectable on stained gels at 0.001% (28 μM) but not
at 0.0003% (8.3 μM). The limit of detection in the assay used by Grossman et al. [6] to test for
possible lactose in their inducing medium was stated by them to be 0.002%, in the range where
induction of P35 protein was observed. I conclude that the unintended induction described by
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 15 of 41
Grossman et al. and observed by us in media made with N-Z-amine from our new barrel is due
to small amounts of lactose in the medium. This seems entirely reasonable, as N-Z-amine is
an enzymatic digest of casein, a milk protein, and milk contains lactose. The casein would have
been purified before digestion, but differing trace amounts of lactose remaining in the final
product presumably account for the differences in inducing activity in different lots of N-Zamine
or tryptone. The finding that glucose prevents unintended induction is also consistent
with a large body of work showing that the presence of glucose in the medium prevents the
uptake and utilization of lactose [16-20]. In retrospect, we were lucky that the barrel of N-Zamine
used for most of our previous work in developing the T7 expression system had low
enough levels of lactose to be free of unintended induction.
Amino acids suppress induction by lactose in log-phase growth—Although the
presence of a small amount of lactose in the medium explains most observations related to
unintended induction, it seemed curious that B834(DE3)P35 could grow to relatively high
density in ZYP containing 0.05-1% lactose, even though high levels of induction kill the cells
(Table 6). Indeed, the titer per A600 indicated that more than 90% of the cells in the saturated
cultures were incapable of forming a colony. Similar results were obtained with B834(DE3)
RIL producing yeast target protein P21, which was used for an extensive exploration of
induction phenomena. Total proteins of cells growing in ZYP containing 0.5% lactose showed
no detectable P21 protein in early log phase but rapid, high-level production as the growth rate
slowed on approach to saturation (Figure 1A), similar to the timing observed by Grossman et
al. [6]. The time course looked similar whether the medium contained 0.1, 0.2, 0.5, 1 or 1.5%
lactose, with induction in each case beginning at A600 ∼1 and reaching a maximum level of
P21 protein per A600 at A600 ∼3, which was maintained to A600 ∼5-6. When incubation was
continued for 15 hours overnight, further increases in culture density were greater the higher
the lactose concentration, reaching as high as A600 ∼14.8 in 1.5% lactose. However, the amount
of target protein per A600 was much reduced (Figure 1A), and titers showed that the density
increases were due primarily to overgrowth of the culture by cells that had lost plasmid. Such
overgrowth can occur in ZYP medium even at the kanamycin concentration of 100 μg/ml used
in these experiments (see section on High phosphate promotes kanamycin resistance, above).
Something in ZYP medium prevents induction by lactose during log-phase growth.
Conceivably, small amounts of glucose or other PTS sugars could be responsible, but N-Zamine
and yeast extract are both rich in amino acids and it seemed possible that amino acids
somehow prevent or modulate the lethal levels of expression that would otherwise be induced
by lactose. P medium containing 1.25% glycerol as a carbon and energy source was used to
test the ability of purified amino acids to allow growth in the presence of 0.1% lactose (Table
7). No growth was apparent in the absence of amino acids, consistent with the inability of
glycerol to prevent lactose induction that is strong enough to prevent cell growth. However,
addition of 18 amino acids, each at a concentration of 100μg/ml, allowed growth to high density
with full induction of P21 protein. Of three subgroups of amino acids, only the group containing
serine supported overnight growth, as did serine itself but not other amino acids in that
subgroup. Although serine seems to be the most effective amino acid in suppressing induction
and allowing growth in the presence of lactose, the combination of 17 amino acids lacking
serine promoted growth in the presence of lactose almost as well as 18 amino acids including
serine. Apparently, something about the uptake and metabolism of amino acids during logphase
growth prevents or modulates lactose induction of target protein sufficiently to allow
cells to grow, but this inhibition is relaxed and full-blown induction occurs upon approach to
saturation.
Metabolic regulation enables auto-induction—The recognition that lactose can induce
production of target protein but is prevented from doing so by compounds that can be depleted
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 16 of 41
during growth opened the possibility of developing media in which target protein is produced
automatically, without the need to monitor growth and add inducer at the proper time. I call
this auto-induction. Ideally, the expression strain would grow in the auto-inducing medium
without expressing target protein until rather high density, when depletion of inhibitory factors
would allow the lactose present in the medium to induce expression, producing high
concentrations of target protein.
Factors that affect the efficiency and reliability of auto-induction in high-density cultures were
examined systematically in B834(DE3) and BL21(DE3), initially testing expression of the
yeast target protein P21 over a wide range of conditions and then expanding to other proteins,
including bacteriophage T7 proteins that are known to be highly toxic to the host bacterium.
The experiments and conclusions are summarized in the following sections.
Carbon and energy sources for high-level production of target protein by autoinduction—As
described in the Complex media part of the section on Minimum nutritional
requirements for growth to high density, growth in ZYP is limited by lack of a carbon and
energy source. Glucose can support growth to high density, but too much glucose prevents
induction by lactose. Lactose itself can support the growth of BL21(DE3), but the initial
products of lactose catabolism are glucose and galactose, and, since BL21 and B834 cannot
use galactose, half of the carbon and energy of lactose is not available. Perhaps more important,
induced T7 RNA polymerase can be so active that most transcription and protein synthesis in
the cell is directed toward target protein [1]. This competition may limit the production of βgalactosidase
and lactose permease, thereby limiting the ability of lactose to serve as a carbon
and energy source for continued production of target protein.
Glycerol supports growth about as well as glucose and does not prevent induction by lactose.
Cultures supplemented with glycerol grow to much higher densities before and after induction
than with lactose as carbon and energy source (for example, compare Figure 1A and B). BL21
(DE3) can grow on other economical carbon and energy sources, including fructose, maltose
and sorbitol (but not sucrose). In limited tests, maltose and sorbitol gave somewhat inconsistent
growth and induction, offering no apparent advantages over glycerol. Therefore, glycerol was
chosen as a carbon and energy source for both fully defined and complex auto-inducing media.
Many combinations of glycerol, glucose, lactose and purified amino acids were tested to
optimize auto-induction and reliability in producing high concentrations of target protein per
volume of culture.
The standard 5052 mixture of 0.5% glycerol, 0.05% glucose, 0.2% lactose has produced
reliable auto-induction of a wide variety of proteins in a range of media and growth conditions
(Table 1). ZYM-5052 or ZYP-5052 is a good choice for the first attempt to express almost any
new target protein. Auto-induced cultures with highly expressed proteins, such as T7 capsid
protein and yeast P21 protein, often attain densities greater than A600 ∼20, more than twice
the density of BL21(DE3) or B834(DE3) themselves grown in the same medium. Microscope
observations of cells from such highly expressing cultures suggested that the induced cells
continued to elongate fairly uniformly, presumably without dividing.
For some target proteins, higher glycerol and/or amino-acid concentrations can produce higher
culture densities and target protein concentrations, if aeration and other media components are
appropriate for maintaining pH. Auto-induced cultures expressing T7 capsid protein have
reached culture densities of A600 >40 in less than 24 hours in ZYP-5052 supplemented to a
total of 2% glycerol and 25 mM succinate in well-aerated cultures (Figure 1C). Comparably
high densities have also been reached in fully defined media supplemented with purified amino
acids that supply carbon and energy.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 17 of 41
Effect of aeration on timing and level of auto-induction of target protein—In
testing the effect of different concentrations of lactose and glycerol on induction of P21 protein
in ZYP medium, a substantial difference was observed in the amount of protein produced in
ZYP containing 1.875% glycerol but no added lactose on two different days. The only obvious
difference between the cultures appeared to be the level of aeration: a standard 0.5 ml of culture
in a 13×100 mm tube reached saturation at A600 ∼13.9 and pH ∼5.6 with a high level of target
protein, but a 5-ml culture in a 125-ml Erlenmeyer flask, reduced to a highly aerated 1.5 ml by
sampling, reached saturation at A600 ∼20.0 and pH ∼6.7 with barely detectable target protein.
To test more systematically how growth and protein production are affected by level of
aeration, a thousand-fold dilution of B834(DE3)RIL/P21 in 80 ml of ZYP containing 0.625%
glycerol but no added lactose was distributed as 0.25, 0.5, 1 or 2 ml samples in 13×100 mm
tubes and 2.5, 5, 10, 20 or ∼39 ml samples in 125-ml Erlenmeyer flasks, which were all grown
at 37°C in the incubator shaker at 325 rpm to provide a fairly wide range of rates of aeration.
The time course of growth and protein production in the Erlenmeyer flasks containing 5 ml or
more of culture was followed by withdrawing approximately 12 samples totaling about 4 ml
from each, which produced a very high aeration rate all the way to saturation for the 5-ml
culture in the 125 ml flask. Two time points and a total volume of approximately 75-215 μl
were sampled from the remaining cultures before saturation. The saturated cultures were also
titered with and without kanamycin to test for viability and plasmid retention. Saturation
densities and pH, relative target protein levels, and titers are given in Table 8.
As shown in Table 8, the level of target protein and viability of saturated cultures varied
tremendously with the rate of aeration: the highest rates of aeration gave no apparent production
of P21 protein or killing of cells and the lowest rates of aeration produced very high levels of
P21 protein and substantial killing of cells. The different cultures whose densities were
measured in the growth phase (not shown) had about the same growth rate to A600 ∼1.0, where
lack of oxygen started to limit growth rate in the cultures with the lowest rates of aeration. The
most highly aerated culture whose growth rate was followed (5 ml reduced to ∼1 ml in a 125ml
flask) maintained a gradually slowing but steady increase in density all the way to saturation
at A600 ∼14.3, with little induction of target protein. The least well-aerated culture whose
growth rate was followed (∼39 ml in a 125-ml flask) began significant production of target
protein by A600 ∼1.5 and had accumulated high levels by A600 ∼3. The doubling time of the
culture was ∼33 min between A600 of 0.1 and 1 but slowed markedly to ∼150 min between
A600 of 3 and 5. In the next 13 hr after reaching A600 ∼5.3, the culture density reached 10.2,
with no apparent decrease in the amount of target protein per A600. At this point, essentially
no cells that carried plasmid were capable of forming a colony, and cells that had lost plasmid
had not yet overgrown the culture. Intermediate rates of aeration gave growth and induction
behavior intermediate between these two extremes. The standard 0.5 ml cultures in 13×100
mm tubes appeared to provide aeration comparable to about 5-10 ml cultures in 125-ml
Erlenmeyer flasks, considering that ∼4 ml of culture was removed from the 10 ml culture to
follow growth rate in this experiment. In this set of cultures, glycerol probably became depleted
at the higher levels of aeration, and, except for the lowest levels of aeration, most cultures
ultimately reached about the same saturation density and pH even though the amounts of target
protein differed markedly.
The failure to produce target protein at the highest rates of aeration in the above experiment
was due to the low concentration of lactose contributed by the N-Z-amine. Table 9 shows the
saturation densities, target protein levels and titers attained at saturation for two sets of cultures
grown in ZYP containing 0.625% glycerol and different concentrations of lactose. In the first
set, 0.5 ml cultures were grown in 13×100 mm tubes, providing the standard, reasonably good
rate of aeration; in the second set, 1.5-ml cultures were grown in 125-ml Erlenmeyer flasks,
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 18 of 41
providing an even higher rate of aeration. In the first set, target protein was highly induced
even in the absence of added lactose. In the more highly aerated second set, little induction of
target protein or cell killing was apparent at 0.001% or less added lactose and only minimal
amounts of target protein or cell killing were apparent between 0.002% and 0.01% lactose.
The typical high levels of target protein and substantial cell killing seen with 0.5 ml in a 13×100
mm tube were produced only at 0.05% lactose or higher. Clearly, the higher the rate of aeration
the more lactose is needed to induce high-level protein production in auto-inducing media. The
concentration of 0.2% lactose chosen for auto-inducing media seems likely to be high enough
to induce full expression of target protein at almost any rate of aeration likely to be encountered
with shaking vessels.
Inclusion of glucose in auto-inducing media and expression of toxic proteins—
Previous workers used lactose to induce the expression of target proteins in T7 expression
strains in fermenters, adding lactose after glucose was depleted [30] or using a fed-batch
fermentation with mixtures of lactose and glucose, which appeared to provide lower rates of
induction and improved solubility of target protein [31]. However, in testing whether mixtures
of glucose and lactose could produce intermediate rates of production in auto-inducing media,
it was clear that the presence of glucose completely prevented induction by lactose and that
production of target protein occurred only after the glucose was depleted. These observations
are in accord with a wealth of previous literature showing that glucose in the medium prevents
lactose from inducing the lac operon [16-20].
Because glucose both prevents lactose from inducing expression of target protein and is
metabolized preferentially during growth, I expected that simply adjusting the concentration
of glucose in media containing an inducing concentration of lactose could allow auto-induction
at any desired density of an actively growing culture. However, the finding that amino acids
and oxygen level modulate the lactose induction of target proteins meant that fine tuning the
culture density at which auto-induction occurred was not straightforward, particularly in rich
media. Media containing amino acids and lactose but no glucose often exhibited a relatively
slow production of target protein well before a rapid, high-level induction that coincided with
slowing growth due to oxygen limitation. However, the presence of glucose always strongly
prevented production of target protein. Therefore, a good approach appeared to be to include
glucose in auto-inducing media at a concentration that would not be depleted until the culture
had grown to moderate density, preferably just before the oxygen depletion that appears to
trigger high-level production of target protein. The effects of different concentrations of
glucose on the level of target protein accumulated were tested in different fully defined and
complex media, in standard 0.5-ml cultures in 13×100 tubes, in time courses with larger
volumes of culture, and at different levels of aeration. A glucose concentration of 0.05%
seemed to be effective over a range of conditions and was selected for inclusion in the autoinducing
media given in Table 1.
An important question is whether the presence of 0.05% glucose completely prevents lactose
from increasing the basal level of target protein in the early stages of growth in auto-inducing
media. When target proteins are highly toxic to the cell, even a small increase in basal
expression over that maintained in non-inducing media might have a significant effect on the
ability of an expression strain to grow and maintain inducible plasmids until auto-induction
takes place. This was tested with clones capable of expressing T7 gene 5.3 and 7.7 proteins,
whose functions are unknown but which are highly toxic to BL21(DE3) and difficult to
maintain and express [3, 4]. Certain plasmid clones capable of expressing 7.7 protein were
toxic enough that BL21(DE3) transformants were not obtained on ZYB plates, which had
inducing activity, but they were obtained on fully defined PAG plates, which lack inducing
activity. These expression strains were stably maintained in PG and MDG non-inducing media,
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 19 of 41
and could be grown and auto-induced in PA-5052, ZYP-5052 and ZYM-5052 media to produce
a strong double band at the approximate position expected for 7.7 protein in electrophoretic
patterns of total cell proteins.
The 5.3 protein is even more toxic to BL21(DE3), and clones capable of expressing it could
be obtained only in vectors specifically modified to accept and express highly toxic proteins
(to be described elsewhere). Again, these expression strains were stable in non-inducing media
and could be grown and induced in auto-inducing media. Auto-induction of active 5.3 protein
caused the culture to stop increasing in density beyond A600 ∼0.5 -1.5, presumably because
of the toxic effect of the target protein on the host. A mutant 5.3 protein having a single aminoacid
substitution produced a relatively strong band at the approximate position expected for
5.3 protein in electrophoretic patterns of total cell proteins. However, wild-type 5.3 protein
was not detected, presumably because protein synthesis stopped before enough 5.3 protein
accumulated to become visible over the background. Clearly, basal expression of target protein
in auto-inducing media containing 0.05% glucose is low enough in the initial stages of growth
that strains capable of expressing target proteins that are highly toxic to BL21(DE3) can be
grown and target protein expressed in auto-inducing media.
Auto-induction is widely applicable and generally superior to IPTG induction for
protein production—Auto-induction is more convenient than IPTG induction because the
expression strain is simply inoculated into auto-inducing medium and grown to saturation
without the need to follow culture growth and add inducer at the proper time. Furthermore, the
culture density and concentration of target protein per volume of culture are typically
considerably higher than what we had been obtaining by IPTG induction. Therefore, even as
the auto-induction phenomenon was being explored and media were being optimized, autoinduction
was being applied with great success to the production of proteins of interest to us
and to colleagues in our department. Nevertheless, a more systematic and wider comparison
of auto-induction and IPTG induction was undertaken.
At hand were expression strains for the first hundred or so yeast proteins selected for our
structural genomics pilot project. The coding sequences had been cloned in pET-13a or
pET-28b, both of which transcribe the target from a T7lac promoter. The expression host was
B834(DE3), with or without the RIL plasmid that supplies tRNAs for codons rarely used by
E. coli. The presence of RIL substantially increased production of several of the yeast target
proteins and did not decrease the production of any. All of these clones had already been tested
for expression and solubility of target proteins by conventional IPTG induction in M9ZYB at
both 37°C and 20°C.
In all, 72 of the yeast clones were tested for expression and solubility of the target protein by
auto-induction, most of them at both 37°C and 20°C, and the results were compared with the
previous IPTG inductions. For 14 clones, IPTG and auto-induction were compared directly in
the same experiment. In general, the level of expression per A600 of culture density and the
solubility of target proteins appeared to be comparable whether expression was induced by
adding IPTG or by auto-induction. The auto-induced cultures typically had considerably higher
densities and therefore also had considerably higher concentrations of target protein per ml of
culture.
Continued incubation of auto-induced cultures for several hours after full induction usually
seemed to have little effect on the solubility or level of target protein per A600 of culture density,
whether the culture was in a medium where the density remained constant or continued to
increase slowly after full induction. This stability of auto-induced cultures at or near saturation,
together with the relative uniformity of the inoculating cultures grown to saturation in nonProtein
Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 20 of 41
inducing media, makes it convenient to screen many strains in parallel for expression and
solubility (or larger cultures for purification) simply by incubating thousand-fold dilutions
overnight at 37°C, or somewhat longer at 20°C. IPTG-induced cultures, on the other hand,
were usually collected 3 hr after induction at 37°C to avoid overgrowth by unproductive cells.
Occasionally, continued incubation for many hours at 37°C reduced the apparent solubility of
the target protein in an IPTG-induced or auto-induced culture or both. In the rare cases where
such behavior was observed, the target protein appeared soluble in a parallel 20°C induction.
It is important to note that auto-induction and saturation often occur at considerably higher
density at 20°C than at 37°C (perhaps due to the higher solubility of oxygen at the lower
temperature). Higher saturation densities combined with slower growth at 20°C means that
cultures may be quite dense after overnight incubation but not yet be induced, so care must be
taken not to collect low-temperature cultures before they have saturated. The incubation time
can be shortened by incubating at 37°C for a few hours, until cultures become lightly turbid,
and then transferring to 20°C for auto-induction.
Auto-induction has become the standard procedure in our laboratory for testing expression and
solubility of proteins produced by T7 expression strains and for producing target proteins in
large amounts for purification. IPTG induction is rarely if ever used anymore. Several yeast
proteins were produced by auto-induction and purified for possible structure determination.
Three that gave crystals suitable for structure determination were labeled with SeMet by autoinduction
(as described in the next section) and yielded structures (P35, PDB 1TXN; P89, PDB
1NJR; and P96, PDB 1NKQ), as has the human SSAT protein (through collaboration with J.
Flanagan and M. Bewley). Dax Fu of this department (personal communication and [32]) has
found that auto-induction increased the yields of five different bacterial integral-membrane
proteins about ten-fold over the previous IPTG induction, to approximately 30-50 mg/liter. So
far, he has determined the structure of one of them. Recipes and protocols for auto-induction
have been distributed to many other laboratories, including structural genomics centers, and
are proving to be highly successful.
Auto-induction for labeling proteins with SeMet for crystallography—Labeling
proteins with SeMet is a standard and useful way to obtain phases for structure determination
by X-ray crystallography [33]. Auto-induction seemed promising as a way to produce SeMetlabeled
target proteins simply and efficiently. Expecting that a methionine-requiring host
would be needed for efficient incorporation of SeMet, I started with B834(DE3), which had a
methionine requirement of unknown genotype [10]. However, it turned out that SeMet labeling
is equally efficient in BL21(DE3).
The concentration of methionine required for growth and auto-induction of B834(DE3)RIL/
P21 was tested in a fully defined auto-inducing medium comparable to PA-5052. Cultures
grew at the normal rate until the methionine was depleted, when the culture density abruptly
stopped increasing. Concentrations equal to or greater than ∼100 μg/ml of methionine were
saturating for growth to A600 ∼10.6. Production of target protein was not apparent at
methionine concentrations less than ∼40 μg/ml (where density stopped increasing at A600
∼3.8) but increased with methionine concentration until the maximum amount of target protein
per A600 was reached at approximately 90 μg/ml. As for complex components, N-Z-amine
supplied saturating amounts of methionine but yeast extract did not. Growth in YP-5052
stopped at A600 ∼3.9 without induction of target protein, equivalent to ∼40 μg/ml of
methionine; addition of 100 μg/ml of methionine to the medium allowed saturation at A600
∼9.5 with full induction. The concentration of methionine in this lot of yeast extract, although
not enough to support good auto-induction, is probably too high to make it a useful supplement
in auto-inducing media for labeling with SeMet.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 21 of 41
Growth and auto-induction in SeMet is stimulated by methionine B834(DE3) expressing
yeast target protein P07, which had previously been labeled with SeMet in the process of
structure determination [13], was used to test the potential for labeling with SeMet in autoinducing
media comparable to PA-5052. Total replacement of methionine by SeMet was not
effective: SeMet concentrations of 50 or 100 μg/ml supported growth relatively poorly, to
A600 of less than 2, with no induction of target protein, and 150 and 200 μg/ml prevented
growth entirely. To test whether small amounts of methionine might alleviate the toxic effects
of SeMet, growth and auto-induction were tested in PA-5052 containing 5-30 μg/ml of
methionine plus 50-200 μg/ml of SeMet. Indeed, cultures containing as much as 150 μg/ml of
SeMet attained A600 ∼5 to ∼8 (considerably higher than with either methionine or SeMet
alone) and induced large amounts of target protein. However, the toxic effects of 200 μg/ml
of SeMet overcame even 30 μg/ml of methionine, reducing saturation density to A600 ∼1.9
and preventing auto-induction. Both SeMet and methionine seem to be used at all stages of
growth, because growth rate in 30 μg/ml methionine was reduced by the presence of 100 μg/
ml SeMet, and the growth curve seemed not to have a discontinuity that might indicate a strong
preferential use of methionine until depletion. The stimulatory effect of methionine on growth
and auto-induction in 100 μg/ml of SeMet was comparable between 10 and 30 μg/ml of
methionine but was significantly diminished in 5 μg/ml.
These results suggested that auto-induction might produce target protein with more than 90%
replacement of methionine by SeMet, if the methionine needed to stimulate growth and autoinduction
could be less than 10% of the amount of SeMet in the medium. To test the level of
incorporation of SeMet into target protein, 100-ml cultures of B834(DE3) expressing Histagged
P07 were grown in defined media comparable to PA-5052, containing either 200 μg/
ml of methionine or 10 μg/ml of methionine plus 100 μg/ml of SeMet. The cultures saturated
at A600 ∼8.8 and 6.7 and yielded 2.8 and 1.9 mg of purified P07 protein, respectively. Mass
spectroscopy determined that the P07 protein from the SeMet-containing culture was more
than 90% labeled with SeMet.
Yeast target protein P89 was only partially soluble but had been purified from auto-induced
cultures and crystallized, so it was a good candidate for SeMet labeling by auto-induction.
However, a test of 0.5 ml culture in the medium used for SeMet labeling of P07 produced rather
small amounts of soluble P89. In an attempt to improve the yield, different concentrations of
methionine and SeMet were tried along with different concentrations of the other 17 amino
acids plus a mixture of 9 vitamins. Interestingly, the vitamins had no effect in PA-5052 itself
but significantly increased both the saturation density and level of P89 produced by autoinduction
in the presence of SeMet. Tests of the individual vitamins showed that vitamin
B12 was the only one needed for the stimulation: a mixture of the other 8 vitamins provided
no stimulation. As little as 3 nM vitamin B12 was sufficient to provide maximum stimulation.
Growth and auto-induction of B834(DE3)RIL/P89 in 400 ml of PASM-5052 (which contains
10 μg/ml methionine, 125 μg/ml SeMet and 100 nM vitamin B12) produced 4 mg of purified
SeMet protein, sufficient for phasing and structure determination [14].
B834 is a metE mutant In control experiments, it was discovered that the presence of vitamin
B12 in the medium allows normal growth of B834(DE3) in the absence of methionine. This
unexpected result shows that the methionine deficiency of B834 is due to a mutation in
metE, which specifies the vitamin B12-independent homocysteine methylase of E. coli, which
catalyzes the last step of methionine synthesis [34]. (Previous reports [35, 36] that B834 is a
metB mutant provided no supporting data and must be incorrect.) E. coli also contains a vitamin
B12-dependent homocysteine methylase, specified by metH. However, since E. coli is
incapable of synthesizing vitamin B12, this enzyme is active only when vitamin B12 is present
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 22 of 41
in the growth medium. Concentrations of vitamin B12 greater than ∼0.75 nM allowed maximal
growth and auto-induction of B834(DE3)RIL/P21 in PA-5052 lacking methionine.
The discovery that B834 is a metE mutant suggests a possible explanation for the stimulation
of growth and auto-induction by vitamin B12 in the presence but not the absence of SeMet.
The presence of methionine in the growth medium represses the synthesis of all of the enzymes
specific for methionine synthesis except for the metH enzyme [34]. SeMet seems likely to have
the same effect, since its presence in the growth medium inhibits the growth of BL21(DE3)
and B834(DE3) to about the same extent even though BL21(DE3) would be fully competent
to synthesize methionine. An important role for methionine is incorporation into Sadenosylmethionine,
a methyl donor in reactions that generate S-adenosylhomocysteine as a
product, which is ultimately metabolized to homocysteine [37]. Since methionine is not toxic,
concentrations in the growth medium can always be made high enough to supply all of the
needs for methionine without recycling homocysteine. However, at the concentrations of
SeMet that can be tolerated in the growth medium, a substantial fraction may ultimately end
up in Se-homocysteine, and the remaining SeMet may be insufficient for continued growth
and synthesis of target proteins. The stimulatory effect of vitamin B12 might be due to its
activation of the metH homocysteine methyltransferase, which regenerates SeMet from Sehomocysteine.
If this interpretation is correct, vitamin B12 might also be expected to stimulate
growth and auto-induction of target proteins in BL21(DE3) growing in the presence of SeMet.
Some stimulation of target protein by the presence of vitamin B12 in PASM-5052 was apparent
in one test with BL21(DE3)P19 but not in a second test. Vitamin B12 is included in PASM-5052
at a concentration of 100 nM.
SeMet labeling in BL21(DE3) The efficient substitution of SeMet for methionine in B834
(DE3) in the presence of vitamin B12 reinforced the conclusion that the combination of 10
μg/ml of methionine and 125 μg/ml of SeMet in PASM-5052 must repress the endogenous
synthesis of methionine. Therefore, SeMet labeling by auto-induction in BL21(DE3), which
does not require methionine for growth, should be just as efficient as in B834(DE3). Indeed,
auto-induction of human spermidine/spermine acetyl transferase (SSAT) in PASM-5052
produced greater than 90% substitution of SeMet for methionine whether produced from B834
(DE3)RIL or BL21-Gold(DE3)RIL. Thus, target proteins can be efficiently labeled with SeMet
by auto-induction in BL21(DE3), and the use of B834(DE3) is not necessary.
Generality of SeMet labeling To explore the general utility of auto-induction for SeMet
labeling, production and solubility of target proteins in PASM-5052 relative to ZYP-5052 and
PA-5052 were tested for 10 different yeast proteins expressed in B834(DE3)RIL by autoinduction
of cultures grown from 1000-fold dilutions at both 37°C and 20°C. (In this set,
PASM-5052 contained 100 μg/ml of SeMet.) All of the 37°C cultures appeared to be saturated
and were sampled for gel electrophoresis after an overnight incubation of 14 hours. The 20°C
cultures in ZYP-5052 and PA-5052 were sampled after 22 hours, but SeMet appears to inhibit
growth much more strongly at 20°C than at 37°C, and the 20°C PASM-5052 cultures were not
sampled until 65 hours and again at 85 hours. All but one of these 10 target proteins appeared
to be produced about as well and to have comparable solubility in PASM-5052 as in the other
two media at both temperatures. These results indicate that auto-induction in PASM-5052
should be generally useful for SeMet labeling of target proteins.
A SeMet concentration of 125 μg/ml was chosen for PASM-5052 medium because it seems
sufficient but not much in excess of the amount needed to support growth and auto-induction
in the presence of 10 μg/ml of methionine and 100 nM vitamin B12. Cultures grown in
significantly higher concentrations of SeMet tended to become an orange brown color upon
prolonged incubation at saturation. The yield of the few SeMet-labeled proteins we have
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 23 of 41
produced by auto-induction in PASM-5052 for structure determination has been comparable
to the yield of the unlabeled proteins.
Auto-induction for labeling proteins with 15N and 13C for NMR—Determination of
protein structures by NMR requires substantial amounts of protein labeled with 15N or 13C.
Auto-induction in fully defined minimal media is potentially very efficient at incorporating
isotopic labels into target protein. Target proteins can be uniformly labeled with 15N simply
by using 15N-labeled (NH4)2SO4 in P-5052 or NH4Cl in N-5052 or LS-5052. Auto-induction
with a well expressed target protein will almost deplete the 50 mM ammonium ion in these
media, so use of the isotope should be very efficient. Reduction below about 25 mM ammonium
will significantly reduce the amount of well-expressed target protein obtained.
Glycerol can be used as a source of 13C for labeling target proteins produced by auto-induction.
Glucose, the most economical source, cannot be used because it prevents auto-induction.
Fortunately, 13C glycerol is relatively economical and undoubtedly would become cheaper if
usage increased. The glucose in the auto-inducing medium will have been depleted by the time
target protein synthesis begins, but lactose metabolism is necessary for auto-induction and
some carbon from lactose is likely to be available for incorporation into target protein, at least
in the early stages of synthesis. The usual auto-induction media contain 0.2% lactose and 0.5%
glycerol, concentrations chosen to ensure maximal production of target protein even at the
highest rates of aeration and to make it unlikely that the culture will go irreversibly acid even
at relatively low rates of aeration. For efficient 13C labeling of target protein, the flow of carbon
from glycerol into target protein should be maximized and the flow from lactose minimized.
As discussed in the section on Effect of aeration on timing and level of auto-induction of target
protein, the concentration of lactose needed for maximal induction of target protein decreases
with decreased rate of aeration. Using a minimal medium containing 100 mM phosphate for
good buffering, auto-induction of T7 capsid protein was followed as a function of decreasing
lactose concentration at the different rates of aeration provided by different volumes of culture
in 13×100 mm tubes. Increasing concentrations of glycerol were also tested. Based on these
tests, the auto-inducing medium C-750501 (Table 1) should provide good 13C labeling of target
protein from glycerol. This medium contains 0.75% glycerol and 0.01% lactose, so almost all
of the carbon entering target protein should be derived from glycerol. High-level induction was
obtained at the aeration rate delivered by 0.75 ml cultures in 13×100 ml tubes, a somewhat
lower aeration rate than with the standard 0.5 ml per tube. Induced cultures saturated at A600
∼10 in less than 24 hours at 37°C with a pH usually above 6.0. These conditions seem likely
to scale to approximately 200-400 ml of culture in an unbaffled 1-liter Erlenmyer flask or
perhaps as much as 1 liter in a 1.8-liter baffled Fernbach flask. Since our structural genomics
project does not involve NMR, I have not tested whether the efficiency of 13C labeling in this
medium is adequate for structure determination. However the recipe has been distributed to
several NMR groups in hopes that it will prove useful. A test of 13C incorporation as a function
of lactose concentration might find that higher lactose concentrations and higher rates of
aeration also provide satisfactory labeling.
Auto-induction with arabinose—Expression systems in which transcription is controlled
by the pBAD promoter of the arabinose operon have relatively low basal expression, which
can make them useful for maintaining and expressing toxic genes [38, 39]. The AraC protein
regulates the pBAD promoter both positively and negatively, and basal expression is further
reduced in the presence of glucose. Expression from the pBAD promoter is induced by
arabinose and modulated by catabolite repression. At least two groups have reported expression
systems in which the sequence for T7 RNA polymerase has been placed under control of the
pBAD promoter [40, 41], and Invitrogen markets BL21-AI, in which the pBAD promoter can
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 24 of 41
express T7 RNA polymerase from the chromosome of BL21. Basal expression of T7 RNA
polymerase from the pBAD promoter in BL21-AI is expected to be lower than that from the
lacUV5 promoter in BL21(DE3), which is probably intrinsically leaky because β-galactosidase
activity is needed to convert lactose to allolactose, the natural inducer of the lactose operon
[42]. Therefore clones expressing highly toxic target proteins from a T7 promoter might be
maintained and expressed more readily in BL21-AI than in BL21(DE3). Indeed, several clones
capable of expressing the highly toxic T7 gene 5.3 protein that could not be established in BL21
(DE3) were readily established in BL21-AI.
When transcription of the target gene is from a T7lac promoter, as in the expression clones we
are using, full expression requires both induction of T7 RNA polymerase and release of lac
repressor from its binding site in the T7lac promoter. In BL21(DE3), both events are triggered
by release of the lac repressor, which is conventionally induced by IPTG or auto-induced by
the presence of lactose in the medium. The combination of expressing T7 RNA polymerase
from a pBAD promoter in the chromosome and the target gene from the T7lac promoter in a
multi-copy plasmid provides enough control that auto-induction of target protein production
is feasible in BL21-AI. Auto-induction of T7 capsid protein in BL21-AI in ZYM-5052 plus
0.05% L-arabinose showed barely detectable capsid protein at A600 ∼2.7, a distinct band that
increased steadily to high levels between A600 ∼4 and ∼10, and an approximately constant
amount per A600 during continued increase in culture density to A600 ∼26 (Figure 4D). Autoinduction
of the toxic T7 gene 5.3 protein stopped growth at A600 ∼1.7 with no 5.3 protein
apparent in gel patterns, consistent with results obtained in BL21(DE3). The presence of 0.05%
glucose in the auto-inducing medium was necessary to allow growth of the 5.3 clone in the
presence of 0.05% arabinose, and the growth was about as rapid as in the absence of arabinose.
Apparently, the presence of glucose is necessary and sufficient to prevent significant induction
by arabinose in BL21-AI in the early stages of growth in the auto-inducing medium. In contrast
to the results obtained in BL21(DE3), the T7 gene 7.7 protein was not apparent above the
background upon auto-induction in BL21-AI.
L-Arabinose concentrations between 0.01% and 0.5% all induced high-level expression of T7
capsid protein in ZYM-5052 (which contains lactose to induce unblocking of the T7lac
promoter). Capsid protein was detectable in the presence of arabinose and absence of lactose,
but much less was produced than in the presence of both, providing a measure of how
effectively bound lac repressor blocks transcription from the T7lac promoter (to be described
elsewhere).
To get a broader comparison of auto-induction in BL21(DE3) and BL21-AI, 42 different clones
of yeast coding sequences under control of the T7lac promoter, which were known to be well
expressed in BL21(DE3)RIL, were also placed in BL21-AI/RIL. In parallel tubes, the BL21
(DE3)RIL clones were auto-induced in ZYM-5052 and the BL21-AI/RIL clones were autoinduced
in ZYM-5052 containing 0.05% L-arabinose. Levels of target protein were generally
comparable in the two hosts, although a few clones appeared to be expressed to a slightly higher
level or to be slightly more soluble in one host or the other. Whether these differences represent
experimental variation or are more significant has not been explored. However, it is clear that
auto-induction from the pBAD and T7lac promoters is generally effective for producing target
proteins in BL21-AI.
Other growth media
High-density cultures for preparation of plasmids—The high-density culture
conditions developed for auto-induction also are convenient for preparation of plasmid DNAs.
Rich media such as ZYM-505 support growth of the plasmid-containing strains we work with
to culture densities of A600 ∼10 or higher when 1.5-2.5 ml of culture is grown in an 18×150
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 25 of 41
mm tube shaken at 300-350 rpm. Lactose is omitted unless auto-induction is desired. The
presence of 0.05% glucose ensures rapid initial growth with little lag. Typically, yields of
plasmid DNA have been several fold greater than obtained in media previously used for this
purpose, and a single 1.5-ml microfuge tube usually provides more plasmid DNA than needed
for most purposes. Adequate aeration ensures growth to high densities, but even moderate
aeration gives high yields. John Dunn of this department (personal communication) is using
auto-inducing media to obtain high yields of a single-copy plasmid that carries an inducible
replication origin under control of a lac promoter.
Commonly used complex media can be deficient in magnesium—As this paper
was being written, I thought to compare culture densities attained in ZYM-505 with terrific
broth (TRB) and 2xYT, rich media commonly used for high-density growths for preparing
plasmid DNAs [15]. All three media contain an enzymatic digest of casein plus yeast extract,
but in different concentrations: 2xYT contains 16 g tryptone, 10 g yeast extract and 5g NaCl
per liter whereas TRB contains 12 g tryptone and 24 g yeast extract, 89 mM phosphate and 4
ml glycerol (= 0.5% w/v) per liter. TRB has several of the same components as ZYM-505 but
a considerably higher concentration of yeast extract. Having previously found that ZY was
deficient in magnesium, I tested 2xYT and TRB as described and also containing 2 mM
MgSO4. The results are shown in Table 10.
The most striking result was that adding magnesium to TRB made from a commercial product
(Gibco/BRL) increased the culture density from A600 ∼3.6 to ∼18.6. The stimulation from
adding magnesium to 2xYT (made from our own barrels of N-Z-amine and yeast extract) was
not as large, increasing from A600 ∼5.7 to ∼8.3. The difference indicated that our N-Z-amine
and yeast extract probably had higher levels of magnesium than the lots used to make the Gibco/
BRL product, and indeed, TRB made from our components produced a considerably higher
A600 ∼12.5, increasing to ∼18.1 with added magnesium. Titration of the TRB from Gibco/
BRL indicated that 0.5 mM MgSO4 was sufficient to saturate the growth requirement. These
results confirm that enzymatic digests of casein or yeast extract are likely to be deficient in
magnesium needed for maximum growth (to different degrees in different lots), and that 1-2
mM magnesium ion should be added to complex media made with these components to ensure
maximum growth.
The 50% higher saturation density in TRB + magnesium (A600 ∼18.5) relative to ZYM-505
(A600 ∼12) is due the greater than 2-fold higher concentration of complex components in TRB.
A similar boost could be achieved more economically by increasing the glycerol concentration
of ZYM-505, perhaps balancing pH by adding succinate or an inexpensive and wellmetabolized
amino acid such as aspartate or glutamate.
Discussion
The phenomenon of unintended induction was sporadic, being found in some lots of complex
media but not others [6]. Furthermore, different portions of the same culture might produce
widely different levels of target protein, depending on the rate of aeration (Table 8). The
realization that lactose is responsible for unintended induction made it possible to develop noninducing
media in which T7 expression strains remain stable and viable all the way to
saturation, and reliable auto-inducing media that produce high-density, fully-induced cultures
completely unattended. This was an iterative process, addressing factors that limit growth to
high-density in batch mode, affect the viability or stability of expression strains, or influence
the level of production of target protein. Interestingly, lack of magnesium limits growth in
typical lots of traditional rich media such as tryptone broth or LB (and newer media such as
terrific broth). With a sufficiency of nutrients, the main limiting factors become maintenance
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 26 of 41
of a pH near neutral and the availability of oxygen as the culture becomes dense enough that
the rate of aeration becomes limiting. The media given in Table 1 have been formulated to give
reliable non-induced cultures and good auto-induction over a range of conditions. We currently
use MDG as the non-inducing medium for growing cultures for freezer stocks or working
stocks, MDAG plates for selecting strains that express highly toxic target proteins, ZYM-5052
for auto-induction, and ZYM-505 for growing high-density cultures for preparing plasmids.
The surprising finding that high phosphate concentration in rich media provides substantial
resistance to kanamycin led to the formulation of lower phosphate media in which metabolic
balancing maintains the pH of the medium near neutral. Growth in glucose or glycerol produces
acid whereas growth in amino acids or tricarboxylic acid cycle intermediates such as succinate,
fumarate, malate or citrate increases the pH. Although the presence of glucose or glycerol limits
the catabolism of these other carbon and energy sources [16-21], appropriate mixtures can
support growth to high densities with only moderate excursions toward acid pH followed by
saturation or extended slow growth close to neutral pH. By increasing the glycerol and aminoacid
concentrations well above those given in Table 1, auto-induction of well-expressed target
proteins has produced culture densities of A600 >50 in shaking batch cultures, comparable to
what has been reported in a fermenter [31]. Potentially, auto-induction could produce proteins
economically on a commercial scale, as high-density cultures fully induced for target protein
can be obtained without complex process controls in media made entirely from inexpensive
components such as mineral salts and mixtures of glucose, glycerol and lactose, supplemented
with fumarate, succinate or glutamate.
Auto-induction depends on mechanisms bacteria use to regulate the use of carbon and energy
sources present in the growth medium. If glucose is present, catabolite repression and inducer
exclusion prevent the uptake of lactose by lactose permease, the product of lacY, thought to be
the only means of lactose uptake in wild-type cells [16-20]. When glucose is depleted, lactose
can be taken up by a small amount of lacY present in uninduced cells and converted to
allolactose, the natural inducer, by β-galactosidase, the product of lacZ [42, 43]. Thus,
induction of the lac operon by lactose should require the presence of at least a small amount
of lactose permease and β-galactosidase in the uninduced cell, and auto-induction should not
be effective with strains that lack either of these activities. Contrary to this expectation,
Grossman et al. [6] observed expression of lacZ from the T7lac promoter in a multi-copy
plasmid upon approach to saturation in BL26(DE3), a derivative of BL21 from which the
lac operon has been deleted [4]. Perhaps changes that occur on approach to saturation make
cells permeable to lactose by some other mechanism [6].
The presence of 0.05% glucose in auto-inducing media blocks induction by lactose in the early
stage of growth so effectively that even strains capable of expressing target proteins highly
toxic to the host cell can grow and maintain functional plasmid until induction. In fact, basal
expression may be low enough that antibiotic might not be needed in the auto-inducing medium
to obtain high-level production of many target proteins.
Having a carbon and energy source other than lactose to support continued growth and
production of target protein after induction enhances high-level production of target proteins
from T7 expression strains. T7 RNA polymerase is so active that induction can direct most
transcription and translation to the target protein [1], which might interfere with full induction
of the ability to metabolize lactose for energy. Glycerol does not interfere with induction of
target protein, and its presence in auto-inducting media more than doubled the yield of target
protein relative to what was obtained with equivalent amounts of lactose as the primary energy
source.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 27 of 41
In the absence of glucose, amino acids appear to modulate or prevent induction of target
proteins by lactose in the early stages of growth, until growth slows as oxygen becomes limiting
upon approach to saturation. Complex mechanisms change metabolism in response to amino
acid availability and oxygen levels [22, 44], but I was not aware that they are known to prevent
lactose utilization. Serine appears to be particularly effective in preventing induction of target
proteins by lactose in log phase but even high concentrations of serine do not prevent induction
as cultures approach saturation. Interestingly, serine is the first amino acid to be depleted during
growth on the mixture of amino acids present in a tryptic digest of casein [24, 25]. Perhaps the
ability to prevent induction is related to a need for higher levels of allolactose to induce
expression from the T7lac promoter in a multi-copy plasmid, because higher than normal levels
of lac repressor are present to ensure saturation of all of the repressor binding sites [2, 4].
Slowing of growth upon oxygen limitation might allow higher levels of allolactose to
accumulate, because of increased uptake of lactose from the medium, decreased catabolism of
allolactose or some other change that promotes induction.
Although developed for expressing target proteins in the IPTG-inducible T7 expression system,
auto-induction could in principle be developed for any expression system in which the elements
driving expression of the target protein are induced by a change in metabolic state that is
brought about by growth of a culture. This could include not only promoters whose induction
is prevented by catabolite repression or inducer exclusion, but also, for example, promoters
activated by approach to saturation, oxygen limitation, or depletion of a compound (such as
methionine) whose synthetic pathway is blocked by its presence in the medium. Simply adding
0.05% L-arabinose to the auto-induction media of Table 1 allows them to be used for producing
target proteins in BL21-AI, where T7 RNA polymerase is expressed from the chromosome by
the arabinoseinducible pBAD promoter. Consistent with the general view that the pBAD
promoter has lower basal expression than the lacUV5 promoter, plasmids expressing
bacteriophage T7 proteins that are highly toxic to the host cell were more easily tolerated in
BL21-AI than in BL21(DE3). Level of expression and solubility of most target proteins tested
were comparable in the two hosts.
Auto-induction has proved to be generally useful for producing a wide range of proteins,
including membrane proteins. The stability and viability of cultures grown in non-inducing
media makes it possible to work with many strains in parallel over a period of weeks. Retransformation
or streaking out cultures for a “fresh” single colony, an unfortunate and tedious
practice in many labs, is almost never necessary for reproducibly expressing high levels of
target protein. Cultures for auto-induction are simply inoculated and grown to saturation, which
is much more convenient than IPTG induction and especially convenient for high throughput
testing of many different target proteins for expression and solubility. The high culture densities
attained by auto-induction produce more target protein per volume of culture than IPTG
induction and also make efficient use of expensive reagents when labeling with SeMet or
isotopes. Auto-induction is convenient, efficient and economical for producing proteins at
almost any scale, from analysis of individual proteins in small laboratories to production of
many different proteins in large projects, and possibly even for production of proteins on a
commercial scale.
Acknowledgements
I am grateful for the enthusiasm and expert technical support of my co-workers and colleagues. Clones for expressing
yeast proteins for structural genomics were constructed and tested by IPTG induction for expression and solubility by
Sue-Ellen Gerchman, with help in the later stages from Eileen Matz, who constructed the clones of T7 and human
proteins. Auto-induction and purification of normal and SeMet-labeled yeast proteins was by Helen Kycia. Nancy
Manning performed innumerable gel electrophoretic analyses of protein expression and solubility in auto-induced
cultures. Structures of yeast proteins were determined by S. Swaminathan, S. Eswaramoorthy and D. Kumaran. Work
on human SSAT was in collaboration with John Flanagan, Maria Bewley and Vito Graziano, who performed the mass
spectroscopy measurements to determine levels of SeMet substitutions. The work was supported by the Office of
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 28 of 41
Biological and Environmental Research of the U.S. Department of Energy and the Protein Structure Initiative of the
National Institute of General Medical Sciences of the National Institutes of Health, as part of the New York Structural
Genomics Research Consortium.
References
1. Studier FW, Moffatt BA. Use of bacteriophage T7 RNA polymerase to direct selective high-level
expression of cloned genes. J Mol Biol 1986;189:113–130. [PubMed: 3537305]
2. Studier FW, Rosenberg AH, Dunn JJ, Dubendorff JW. Use of T7 RNA polymerase to direct expression
of cloned genes. Methods Enzymol 1990;185:60–89. [PubMed: 2199796]
3. Studier FW. Use of bacteriophage T7 lysozyme to improve an inducible T7 expression system. J Mol
Biol 1991;219:37–44. [PubMed: 2023259]
4. Dubendorff JW, Studier FW. Controlling basal expression in an inducible T7 expression system by
blocking the target T7 promoter with lac repressor. J Mol Biol 1991;219:45–59. [PubMed: 1902522]
5. Kelley KC, Huestis KJ, Austen DA, Sanderson CT, Donoghue MA, Stickel SK, Kawasaki ES, Osburne
MS. Regulation of sCD4-183 gene expression from phage-T7-based vectors in Escherichia coli. Gene
1995;156:33–36. [PubMed: 7737513]
6. Grossman TH, Kawasaki ES, Punreddy SR, Osburne MS. Spontaneous cAMP-dependent derepression
of gene expression in stationary phase plays a role in recombinant expression instability. Gene
1998;209:95–103. [PubMed: 9524234]
7. Giordano TJ, Deuschle U, Bujard H, McAllister WT. Regulation of coliphage T3 and T7 RNA
polymerases by the lac repressor-operator system. Gene 1989;84:209–219. [PubMed: 2693210]
8. Lopez PJ, Guillerez J, Sousa R, Dreyfus M. On the mechanism of inhibition of phage T7 RNA
polymerase by lac repressor. J Mol Biol 1998;276:861–875. [PubMed: 9566192]
9. Burley SK, Almo SC, Bonanno JB, Capel M, Chance MR, Gaasterland T, Lin D, Sali A, Studier FW,
Swaminathan S. Structural genomics: beyond the human genome project. Nat Genet 1999;23:151–
157. [PubMed: 10508510]
10. Wood WB. Host specificity of DNA produced by Escherichia coli: bacterial mutations affecting the
restriction and modification of DNA. J Mol Biol 1966;16:118–133. [PubMed: 5331240]
11. Rosenberg AH, Lade BN, Chui DS, Lin SW, Dunn JJ, Studier FW. Vectors for selective expression
of cloned DNAs by T7 RNA polymerase. Gene 1987;56:125–135. [PubMed: 3315856]
12. Gerchman SE, Graziano V, Ramakrishnan V. Expression of chicken linker histones in E. coli: sources
of problems and methods for overcoming some of the difficulties. Protein Expr Purif 1994;5:242–
251. [PubMed: 7950367]
13. Eswaramoorthy S, Gerchman S, Graziano V, Kycia H, Studier FW, Swaminathan S. Structure of a
yeast hypothetical protein selected by a structural genomics approach. Acta Crystallogr D Biol
Crystallogr 2003;59:127–135. [PubMed: 12499548]
14. Kumaran D, Eswaramoorthy S, Studier FW, Swaminathan S. Structure and mechanism of ADPribose-1“-monophosphatase
(Appr-1”-pase), a ubiquitous cellular processing enzyme. Protein
Science in press. 2005
15. Sambrook, J.; Fritsch, EF.; Maniatis, T. Molecular Cloning: A Laboratory Manual. 2nd. Cold Spring
Harbor Laboratory Press; Cold Spring Harbor, New York: 1989.
16. Meadow ND, Fox DK, Roseman S. The bacterial phosphoenolpyruvate: glycose phosphotransferase
system. Annu Rev Biochem 1990;59:497–542. [PubMed: 2197982]
17. Postma, PW.; Lengeler, JW.; Jacobson, GR. Phosphoenolpyruvate: Carbohydrate Phosphotransferase
Systems. In: Niedhardt, FC., editor. Eschericha coli and Salmonella Typhimurium: Cellular and
Molecular Biology. 2nd. American Society for Microbiology; Washington, D.C.: 1996. Chapter 75
18. Saier, MH., Jr.; Ramseier, TM.; Reizer, J. Regulation of Carbon Utilization. In: Neidhardt, FC., editor.
Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology. 2nd. American
Society for Microbiology; Washington, D.C.: 1996. Chapter 85
19. Inada T, Kimata K, Aiba H. Mechanism responsible for glucose-lactose diauxie in Escherichia coli:
challenge to the cAMP model. Genes Cells 1996;1:293–301. [PubMed: 9133663]
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 29 of 41
20. Kimata K, Takahashi H, Inada T, Postma P, Aiba H. cAMP receptor protein-cAMP plays a crucial
role in glucose-lactose diauxie by activating the major glucose transporter gene in Escherichia coli.
Proc Natl Acad Sci U S A 1997;94:12914–12919. [PubMed: 9371775]
21. Eppler T, Postma P, Schutz A, Volker U, Boos W. Glycerol-3-phosphate-induced catabolite
repression in Escherichia coli. J Bacteriol 2002;184:3044–3052. [PubMed: 12003946]
22. Lynch, AS.; Lin, ECC. Responses to Molecular Oxygen. In: Neidhardt, FC., editor. Escherichia coli
and Salmonella typhimurium: Cellular and Molecular Biology. 2nd. American Society for
Microbiology; Washington D.C.: 1996. Chapter 95
23. Holms H. Flux analysis and control of the central metabolic pathways in Escherichia coli. FEMS
Microbiol Rev 1996;19:85–116. [PubMed: 8988566]
24. Chang DE, Shin S, Rhee JS, Pan JG. Acetate metabolism in a pta mutant of Escherichia coli W3110:
importance of maintaining acetyl coenzyme A flux for growth and survival. J Bacteriol
1999;181:6656–6663. [PubMed: 10542166]
25. Kumari S, Beatty CM, Browning DF, Busby SJ, Simel EJ, Hovel-Miner G, Wolfe AJ. Regulation of
acetyl coenzyme A synthetase in Escherichia coli. J Bacteriol 2000;182:4173–4179. [PubMed:
10894724]
26. Slonczewski, JL.; Foster, JW. pH-Regulated Genes and Survival at Extreme pH. Escherichia coli and
Salmonella typhimurium: Cellular and Molecular Biology. 2nd. Neidhardt, FC., editor. American
Society for Microbiology; Washington, D.C.: 1996. Chapter 96
27. McFall, E.; Newman, EB. Amino Acids as Carbon Sources. In: Neidhardt, FC., editor. Escherichia
coli and Salmonella typhimurium: Cellular and Molecular Biology. 2nd. American Society for
Microbiology; Washington, D.C.: 1996. Chapter 22
28. Neidhardt, FC. Biosynthesis of Amino Acids and Nucleotides. 2nd. American Society for
Microbiology; Washington, D.C.: 1996. Escherichia coli and Salmonella typhimurium: Cellular and
Molecular Biology. Section B1
29. Wanner, BL. Phosphorus Assimilation and Control of the Phosphate Regulon. In: Neidhardt, FC.,
editor. Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology. 2nd.
American Society for Microbiology; Washington, D.C.: 1996. Chapter 87
30. Neubauer P, Hofmann K, Holst O, Mattiasson B, Kruschke P. Maximizing the expression of a
recombinant gene in Escherichia coli by manipulation of induction time using lactose as inducer.
Appl Microbiol Biotechnol 1992;36:739–744. [PubMed: 1369364]
31. Hoffman BJ, Broadwater JA, Johnson P, Harper J, Fox BG, Kenealy WR. Lactose fed-batch
overexpression of recombinant metalloproteins in Escherichia coli BL21 (DE3): process control
yielding high levels of metal-incorporated, soluble protein. Protein Expr Purif 1995;6:646–654.
[PubMed: 8535158]
32. Chao Y, Fu D. Thermodynamic studies of the mechanism of metal binding to the Escherichia coli
zinc transporter YiiP. J Biol Chem 2004;279:17173–17180. [PubMed: 14960568]
33. Hendrickson WA, Horton JR, LeMaster DM. Selenomethionyl proteins produced for analysis by
multiwavelength anomalous diffraction (MAD): a vehicle for direct determination of threedimensional
structure. Embo J 1990;9:1665–1672. [PubMed: 2184035]
34. Cohen, GN.; Saint-Girons, I. Biosynthesis of threonine, lysine, and methionine. In: Neidhardt, FC.,
editor. Escherichia coli and Salmonella Typhimurium: Cellular and Molecular Biology. American
Society for Microbiology; Washington, D.C.: 1987. p. 429.-444.
35. Budisa N, Steipe B, Demange P, Eckerskorn C, Kellermann J, Huber R. High-level biosynthetic
substitution of methionine in proteins by its analogs 2-aminohexanoic acid, selenomethionine,
telluromethionine and ethionine in Escherichia coli. Eur J Biochem 1995;230:788–796. [PubMed:
7607253]
36. Pieper M, Betz M, Budisa N, Gomis-Ruth FX, Bode W, Tschesche H. Expression, purification,
characterization, and X-ray analysis of selenomethionine 215 variant of leukocyte collagenase. J
Protein Chem 1997;16:637–650. [PubMed: 9263126]
37. Miller CH, Duerre JA. S-ribosylhomocysteine cleavage enzyme from Escherichia coli. J Biol Chem
1968;243:92–97. [PubMed: 4867478]
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 30 of 41
38. Cagnon C, Valverde V, Masson JM. A new family of sugar-inducible expression vectors for
Escherichia coli. Protein Eng 1991;4:843–847. [PubMed: 1798708]
39. Guzman LM, Belin D, Carson MJ, Beckwith J. Tight regulation, modulation, and high-level
expression by vectors containing the arabinose PBAD promoter. J Bacteriol 1995;177:4121–4130.
[PubMed: 7608087]
40. Wycuff DR, Matthews KS. Generation of an AraC-araBAD promoter-regulated T7 expression
system. Anal Biochem 2000;277:67–73. [PubMed: 10610690]
41. Chao YP, Chiang CJ, Hung WB. Stringent regulation and high-level expression of heterologous genes
in Escherichia coli using T7 system controllable by the araBAD promoter. Biotechnol Prog
2002;18:394–400. [PubMed: 11934312]
42. Beckwith, J. The Operon: An Historical Account. In: Neidhardt, FC., editor. Escherichia coli and
Salmonella typhimurium: Cellular and Molecular Biology. American Society for Microbiology;
Washington, D.C.: 1987. p. 1439.-1452.
43. Huber RE, Kurz G, Wallenfels K. A quantitation of the factors which affect the hydrolase and
transgalactosylase activities of beta-galactosidase (E. coli) on lactose. Biochemistry 1976;15:1994–
2001. [PubMed: 5122]
44. Newman, EB.; Lin, RT.; D'Ari, R. The Leucine/Lrp Regulon. In: Neidhardt, FC., editor. Escherichia
coli and Salmonella typhimurium: Cellular and Molecular Biology. 2nd. American Society for
Microbiology; Washington, D.C.: 1996. Chapter 94
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 31 of 41
Figure 1.
Electrophoretic patterns of total cell proteins during growth of auto-inducing cultures at 37°C.
Equal culture densities were analyzed in each lane of a set, and the A600 of the culture at the
time of sampling is given above each lane. A) BL21(DE3)RIL/P21 was grown in 6 ml of ZYP
+ 0.5% lactose in a 125-ml Erlenmeyer flask. The culture was sampled every 30 min, except
that the interval before the last sample was 15 hours. The cell suspensions before processing
for electrophoresis were A600 ∼10. B) BL21(DE3)RIL/P21 was grown in 5 ml of ZYP-5052
in a 125-ml flask (except that the glycerol concentration was 0.625% instead of 0.5%). The
culture was sampled every 30 min. The cell suspensions were A600 ∼10. C) BL21(DE3)
T7-10A was grown in 2.5 ml of ZYP-20052 + 25 mM succinate in a 125-ml flask (the glycerol
concentration was 2 %). The culture was sampled every 30-40 min until A600 ∼22.6, and then
intervals of 70 min, 55 min, and 13.5 hours. The cell suspensions were A600 ∼5. D) BL21-AI/
T7-10A was grown in 2 ml of ZYM-5052 + 0.05% L-arabinose in an 18×150 mm culture tube.
The culture was sampled every 30 min until A600 ∼10.9, then three intervals of 60 min and a
final interval of 16 hours. The cell suspensions were A600 ∼2.5.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 32 of 41Table1
auto-inducingmedia.AnexplanationofthenamingconventionsisgivenunderGrowthmediain
growingfreezerstocksandworkingcultures,ZYM-5052forroutineauto-induction,andZYM-505
Comment%N-
Z-
amine
%YeastextractmM
Na2
HPO
4
mM
KH
2PO
4
mM
NH
4Cl
mM
(NH
4)
2SO
4
mM
Na2
SO4
mM
MgSO
4
a
Tracemetal
sb
%Glycerol%Glucose%LactosemMSuccinate%Aspartate%18
amino
acids
c
50502520.2x0.5
50502520.2x0.50.36
15
Nlabeling50502520.2x0.50.050.2
50502520.2x0.50.050.20.36
SeMetlabeling50502520.2x0.50.050.2SeMe
td
10.550502520.2x0.50.050.2
12.512.550520.2x0.520
15
Nlabeling12.512.550520.2x0.50.050.220
252550520.2x0.50.25
252550520.2x0.50.250.36
252550520.2x0.50.050.20.250.36
Plasmidpreps10.5252550520.2x0.50.05
10.5252550520.2x0.50.050.2
15
Nlabeling505050520.2x0.50.050.2
13
Clabeling505050520.2x0.750.050.01
inpreviouslydistributedrecipes
lbutrecommendedinmediacontainingZY
e)
(noC,Y,M),10μg/mlofmethionine,125μg/mlSeMet,and100nMvitaminB12
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 33 of 41
Table 2
Saturation densities and induction of T7 RNA polymerase in different growth media. BL21
(DE3) was grown 17 hr, 37°C from 104 dilution, 10 ml in 125-ml Erlenmeyer flasks. N-Zamine
in growth media was from the new barrel, which had inducing activity, except that the
“Old ZB” culture and the plates for testing 4107 plaque formation were made from the old
barrel, which lacked inducing activity.
T7 4107 plaques a
Growth medium Addition A600 pH Number Size Time of appearance
ZB 1.2 8.25
ZYB 2.8 7.65
3xZYB 7.6 7.35
Old ZB 1.0 7.90  39 small   4.5 hr
ZB IPTG in plate 188 large   2.0 hr
0.5xZBb 0.5 8.62 144 variable   3.0 hr
ZB 1.1 8.22 150 variable   3.0 hr
2xZBb 2.4 8.53 189 variable   2.5 hr
4xZBb 5.4 8.32 183 variable   2.0 hr
8xZBb 6.9 7.92 222 large   1.5 hr
ZB 0.1% glucose 1.3 7.58 150 variable   2.5 hr
ZB 1% glucose 1.2 5.12  38 tiny   4.5 hr
4xZBb 6.0 8.05 231 large   1.5 hr
4xZBb 1% glucose 4.3 5.27  65 tiny, turbid   3.5 hr
8xZBb 5.5 7.37 220 large   1.5 hr
8xZBb 1% glucose 5.8 5.28  45 tiny, turbid   3.5 hr
ZYB 2.5 8.55 175 small    3 hr
ZYB 1% glucose 3.2 5.30  22 small overnight
M9 2.5 6.10
M9ZB 7.5 7.02
M9ZB 2% glucose (total) 5.8 4.57
ZB M9 PO4+NH4Cl 1.2 7.36
ZB 1 mM MgSO4 2.1 8.35
ZB M9 salts 2.6 7.40
ZB 1% glucose 1.2 5.12
ZYc 1.7
ZYc 1 mM MgSO4 4.3
a
Equivalent numbers of T7 4107 deletion phage particles were plated on 0.25 ml of each culture in 2.5 ml of ZB top agar on ZB plates, both made with
N-Z-amine lacking inducing activity (old ZB)
b
Contained 0.5% NaCl
c
Grown 14 hr, 37°C, 0.5 ml in 13×100 mm tubes
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 34 of 41
Table 3
Growth requirements for BL21(DE3) in modified LG media. Cultures were grown 14-15 hr,
37°C from 103 dilution, 0.5 ml in 13×100 mm tubes. Cultures were titered after 3-4 weeks in
the refrigerator.
LG medium, no succinate LG medium + 25 mM succinate
Glucose A600 pH Titer (× 109
) A600 pH Titer (× 109
)
0 0 0
0.05% 0.24 6.89 0.66 7.03  1.8
0.10% 0.7 6.76 1.4 7.20  3.6
0.15% 1.2 6.62 2.2 7.33  7.2
0.20% 1.6 6.51 3.1 7.47 11.8
0.25% 1.8 6.42  5.0 3.2 7.47 15.5
0.30% 2.6 6.19  7.1 4.0 7.64 14.5
0.35% 3.0 5.91  8.1 4.2 7.76 12.0
0.40% 3.0 4.85 4.6 7.69  8.8
0.45% 3.8 4.57 5.1 7.87  7.8
0.50% 3.3 4.41 <0.02 5.3 7.84  9.5
Na2SO4
0 0.7 6.67  0.11 0.6 6.67  0.46
0.1 mM 2.0 6.06 ∼0.14 1.8 6.89  2.1
0.2 mM 2.6 5.50 <0.02 3.2 6.49  3.9
0.5 mM 3.8 4.91 <0.02 6.1 6.75 13.4
1 mM 3.8 4.86 5.9 6.74 12.7
2 mM 3.6 4.84 5.8 6.75 14.2
NH4Cl
0 0 6.98 <0.02 0 7.02
5 mM 1.5 6.51 <0.04 1.5 6.75 ∼0.08
10 mM 3.0 5.50 <0.04 2.8 6.50  4.7
15 mM 3.4 4.90 4.0 5.95  9.1
20 mM 3.4 4.87 4.9 6.52 16.4
25 mM 3.4 4.87 5.3 6.92 12.5
50 mM 3.5 4.90 5.5 7.10 13.4
Phosphate
0 0.78 6.75
1 mM 3.8 6.39
2 mM 3.9 6.09
5 mM 5.1 7.75
10 mM 5.7 7.96
15 mM 5.8 8.15
20 mM 5.9 8.19
25 mM 5.8 8.11
35 mM 5.5 7.83
50 mM 5.6 7.15
MgSO4
0 0 6.90 0 6.79
0.1 mM 0.41 6.81  0.12 3.9 6.73  5.0
0.2 mM 1.1 6.61  0.42 5.7 6.28  5.1
0.5 mM 3.7 5.64 <0.02 6.4 6.20  8.8
1 mM 3.7 5.08 <0.02 6.1 6.51 13.2
2 mM 3.7 4.81 <0.02 5.8 6.80 12.3
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 35 of 41
Table 4
Effectiveness of individual amino acids in balancing pH from 0.5% glucose or in serving as a
carbon and energy source. BL21(DE3) was grown from 103 dilution, 0.5 ml in 13×100 mm
tubes.
Grown in LG medium Grown in L medium + 0.18% aa a
Addition Conc. A600 pH Concentration A600 pH
8 hr 22 hr 5.7 hr 46 hr
Glucose 0.5% 0.58 4.0 4.06 0.5% 28 mM 4.6 8.2 4.28
Glycerol 0.5% 54 mM 2.4 9.0 5.72
Succinate 20 mM 0.74 7.6 6.80 0.5% 42 mM 0.80 4.5 8.80
18 aa 0.27% 6.2 6.9 5.66 0.18% 24 mM 0.56 1.4 6.78
D 0.25% 4.8 8.1 7.55 0.5% 38 mM 0.84 3.9 8.28
S 0.25% 0.12 6.9 6.95 0.5% 48 mM 1.3 5.9 7.43
S + 100
ILV b
0.25% 0.56 8.7 6.93
N 0.25% 2.6 7.0 6.86 0.5% 33 mM 0.72 2.0 7.02
G 0.25% 1.1 6.4 6.58 0.5% 67 mM 0.32 1.0 7.43
E 0.25% 3.4 7.5 6.28 0.5% 30 mM 0.72 5.1 7.28
A 0.25% 0.16 4.2 4.53 0.5% 56 mM 0.60 5.1 7.21
A 0.50% 6.2 7.32
P 0.25% 2.0 5.3 3.63 0.5% 43 mM 0.92 9.6 7.00
T 0.25% 0.52 4.9 3.64 0.5% 42 mM 0.76 3.0 7.09
Q 0.25% 1.3 6.1 3.94 0.5% 34 mM 0.78 1.9 6.71
I 0.25% 1.2 4.3 3.86 0.5% 38 mM 0.44 1.1 6.67
L 0.25% 0.22 3.7 3.65 0.5% 38 mM 0.74 1.5 6.70
V 0.25% 0.26 3.5 3.70 0.5% 43 mM 0.64 1.3 6.73
M 0.25% 0.66 4.3 3.82 0.5% 34 mM 0.72 1.4 6.74
R 0.25% 0.88 4.9 3.63 0.5% 24 mM 0.70 1.5 6.81
K 0.25% 0.90 4.3 3.60 0.5% 27 mM 0.70 1.3 6.75
F 0.25% 0.64 2.3 3.89 0.5% 30 mM 0.68 1.2 6.81
W 0.25% 1.3 4.4 3.68 0.5% 24 mM
Hc 0.25% 0.16 4.5 3.69 0.5% 24 mM 0.06 1.0 5.19
a
100 μg/ml of each of 18 amino acids (no C or Y)
b
0.25% serine plus 100 μg/ml each of isoleucine, leucine and valine
c
Reconstitution indicated an initial pH ∼5.4 for 0.25% histidine in LG and an initial pH ∼6.0 for 0.5% histidine in L + 0.18% amino acids
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 36 of 41
Table 5
Effects of metal ions on saturation density and auto-induction. B832(DE3)RIL/P21 was grown
20 hr, 37°C from 103 dilution, 0.5 ml in 13×100 mm tubes.
Mediuma Addition A600 at different metal concentrations Target protein
Concentration of trace metal mixb
0 0.1x 1x 10x
ZYP-5052 18.0 +++
PA-5052  4.4 (+)
PA-5052 metal mixb 11.3 12.9 13.1 +++
Metal ion concentration
1 uM 10 uM 100 uM
PA-5052 FeCl3  7.8 12.7 13.7 +++
PA-5052 MnCl2 11.8 13.2 12.9 +++
PA-5052 CoCl2 11.1 13.6c ∼0.1d +++
PA-5052 ZnSO4  6.9  8.1  8.7 ++
PA-5052 NiCl2  4.8  7.7  5.2 +
PA-5052 Na2MoO4  7.6  5.5  6.6 +
PA-5052 CaCl2  6.0  4.7  5.3 (+)
PA-5052 CuCl2  5.6  5.2  4.5 (+)
PA-5052 Na2SeO3  5.8  6.3 ∼0.7d (+)
PA-5052 H3BO3  6.1  4.6  5.3 (+)
a
The media contained 0.625% glycerol rather than the usual 0.5% in 5052. The PA medium contained 200 μg/ml of methionine and 100 μg/m of the other
17 amino acids (no C or Y). Both ZYP and PA media also contained 1 μM each of nicotinic acid, pyridoxine, thiamine, vitamin B12, biotin, rboflavin
and folic acid.
b
The trace metal mix differed from the final formulation. This trace metal mix contained 20 μM CaCl2, 10 μM each of FeCl3, MnCl2, and ZnSO4, 0.1
μM CoCl2, and 0.05 μM each of CuCl2 and NiCl2.
c
Growth rate was normal after a lag of about an hour.
d
These cultures had very slow growth.
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 37 of 41
Table 6
Induction as a function of lactose concentration in ZYP made with N-Z-amine that has no
inducing activity. B834(DE3)P35 was grown 13 hr, 37°C from 104 dilution, 0.5 ml in 13×100
mm tubes.
Lactose concentration Target protein
Titer (× 109
) a
Medium A600 plasmidb totalc
ZYP 0  0 5.8 0 12.6  14.0
ZYP 0.005%  0.14 mM 5.8 +++ 10.6  10.7
ZYP 0.01 %  0.28 mM 5.5 +++  6.9   6.7
ZYP 0.02 %  0.56 mM 4.9 +++  4.7   4.6
ZYP 0.05 %  1.4  mM 4.2 +++  ∼0.12   0.48
ZYP 0.1  %  2.8  mM 4.4 +++  ∼0.04  ∼0.16
ZYP 1   % 28   mM 4.1 +++  <0.04   0.9
a
“∼” indicates that titer was based on fewer than 10 colonies
b
Titer of cells that are resistant to kanamycin and therefore retain plasmid
c
Titer in the absence of antibiotic, which includes cells with or without plasmid
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 38 of 41
Table 7
Ability of amino acids to suppress lactose induction and allow growth of B834(DE3)RIL/P21
in P + 1.25% glycerol + 0.1% lactose + 100 μg/ml of methionine. Cultures were grown from
103 dilution, 0.5 ml in 13×100 mm tubes.
Addition (100 μg/ml each) A600 pH Target protein
19 hr, 37°C
   0 0
   20 aa 12.4 6.28 +++
   19 aa (no C) 12.7 5.18 +++
   18 aa (no C,Y) 13.2 5.19 +++
   GACPTKR 0
   ILVSHNQ 11.3 6.35 +++
   FYWDE 0
   ILV 0
   S 4.7 6.69 +++
   H 0
   N 0
   Q 0
14.5 hr, 37°C
   18 aa (no C,Y) 10.6 6.20 +++
   17 aa (no S, C, Y) 8.6 6.33 +++
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 39 of 41
Table 8
Effect of aeration on saturation density and protein production. B834(DE3)RIL/P21 was grown
23 hr, 37°C from 103 dilution in ZYP + 0.625% glycerol without added lactose.
Culture volume (ml)
Saturation
Target protein
Titer (× 109
) a
Vessel A600 pH plasmidb totalc
13×100 mm tube  0.25 15.0 + 21 29
13×100 mm tube  0.5 15.4 7.15 ++ 11 12
13×100 mm tube  1 15.0 7.19 +++  2.6  6.5
13×100 mm tube  2  8.5 6.34 +++  2.2  2.5
125 ml flask  2.5 13.0 7.04 0 25 25
125 ml flask  5 14.8 6.99 0 20 22
125 ml flask  10 14.0 7.08 ++ 15 17
125 ml flask  20 14.3 7.09 +++  2.6  6.4
125 ml flask ∼39  10.2 6.60 +++  ∼0.1  3.4
a
“∼” indicates that titer was based on fewer than 10 colonies
b
Titer of cells that are resistant to kanamycin and therefore retain plasmid
c
Titer in the absence of antibiotic, which includes cells with or without plasmid
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 40 of 41
Table 9
Lactose concentration needed for induction as a function of rate of aeration. B834(DE3)RIL/
P21 was grown 17 hr, 37°C from 104 dilution in ZYP + 0.625% glycerol.
Moderate aeration 0.5 ml in 13×100
mm tube
High aeration 1.5 ml in 125-ml Erlenmeyer flask
Target protein Target protein Titer (× 109
)
Lactose concentration A600 pH A600 pH plasmi
da
tota
lb
0  0 15.3 7.19 +++ 15.3 6.94 ?  12 14
0.0001%  
2.8 
μM
13.8 6.96 ?
0.0002%  
5.6 
μM
13.9 6.97 ?
0.0005% 14 
  
μM
14.0 6.98 ?
0.001 % 28 
  
μM
14.2 6.98 ? 20 21
0.002 % 56 
  
μM
13.5 7.00 (+) 21 18
0.005 %  
0.14 
mM
13.9 7.00 + 21
0.01 %  
0.28 
mM
16.2 7.19 +++ 15.1 6.98 + 17 16
0.02 %  
0.56 
mM
16.4 7.20 +++ 16.9 6.99 ++ 10 10
0.05 %  
1.4 
mM
16.4 7.20 +++ 18.9 7.01 +++  3.3  
6.2
0.1  %  
2.8 
mM
16.6 7.19 +++ 17.9 7.02 +++  2.2  
5.0
0.2  %  
5.6 
mM
16.6 6.93 +++ 19.2 7.00 +++  1.5  
4.1
0.5  % 14 
  
mM
17.1 6.93 +++ 20.4 6.94 +++  0.5  
5.4
1   % 28 
  
mM
18.0 6.33 +++ 27.4 6.82 +++  <0.1  
7.7
a
Titer of cells that are resistant to kanamycin and therefore retain plasmid
b
Titer in the absence of antibiotic, which includes cells with or without plasmid
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.
NIH-PAAuthorManuscriptNIH-PAAuthorManuscriptNIH-PAAuthorManuscript
Studier Page 41 of 41
Table 10
Effect of magnesium on saturation density in 2xYT and terrific broth (TRB). BL21(DE3) was
grown 15 hr, 37°C from 103 dilution, 0.5 ml in 13×100 mm tubes.
Growth medium Source A600 pH
ZYM-505 local 12.0 7.05
2xYT local  5.7 8.37
2xYT + 2 mM MgSO4 local  8.3 8.44
TRB Gibco/BRL  3.6 7.73
TRB + 2 mM MgSO4 Gibco/BRL 18.6 8.21
TRB local 12.5 8.06
TRB + 2 mM MgSO4 local 18.1 8.18
Protein Expr Purif. Author manuscript; available in PMC 2006 May 8.