Immunohistochemistry (IHC) and immunocytochemistry (ICC) are routinely employed for the microscopic identification and diagnosis of cancerous cells in histological tissues and cell cultures. The maximally attainable contrast of conventional histological staining techniques, however, is low. While the anti-Stokes emission of photon-upconversion nanoparticles (UCNP) can efficiently eliminate optical background interference, excluding non-specific interactions of the label with the histological sample is equally important for specific immunolabeling. To address both requirements, we have designed and characterized several UCNP-based nanoconjugates as labels for the highly specific detection of the cancer biomarker HER2 on various breast cancer cell lines. An optimized streptavidin-PEG-neridronate-UCNP conjugate provided an unsurpassed signal-to-background ratio of 319, which was 50-fold better than conventional fluorescent labeling under the same experimental conditions. In combination, the absence of optical interference and non-specific binding lays the foundation for computer-based data evaluation in digital pathology.