Morphological changes were among the first reported phenotypes of senescent cells [1
]. Changes in cell shape, volume and macromolecular content (DNA, RNA, protein) were actively studied in the 1970's but have received little attention since. Unfortunately, some of the original literature is not readily accessible, such as the increase in total nuclear protein of senescent HDF, which is noted in the CRC Handbook of Cell Biology of Aging [34
], but the original reference [29
] is not listed on PubMed. Although the increased protein content of senescent cells in vitro is in general well accepted, whether this occurs in vivo is not known. Our objective was to investigate changes in total protein as a potential new biomarker of aging, but this necessitated first the development of a fluorescence microscopy method to detect and quantify any changes.
While the Molecular Probes Handbook of Fluorescent Probes and Labeling Technologies lists numerous reagents and procedures for the staining of proteins in solution, in gels, on blots, and even on peptide microarrays, not a single application is listed for microscopy. We were attracted to fluorescent dyes because of their greater sensitivity and dynamic range relative to absorbance-based techniques. Two general principles of fluorescence-based protein detection have been developed: reactive dyes that couple with protein amines to form fluorescent, covalent adducts, and dyes that interact non-covalently with hydrophobic regions of proteins or at a protein-detergent interface.
Dyes that couple covalently with primary amines have been widely used, including the labeling of intact or permeabilized cells in flow cytometry [35
], but their compatibility with paraformaldehyde fixation is problematic. This is because paraformaldehyde also reacts with amines, and thus removes the reactive groups needed for subsequent dye coupling. We tested three commonly used amine-reactive dyes (fluorescamine [37
], 3-(4-carboxybenzoyl) quinoline-2-carboxaldehyde (CBQCA) [38
], and ophthaldialdehyde (OPA) [39
]) in conjunction with a variety of organic solvent-based coagulative fixatives (such as methanol, acetone, etc.), but were unable to develop a satisfactory method that provided both good preservation of morphology and consistent, strong staining. An additional issue with amine-reactive dyes is that they show wide protein to protein differences in staining.
Dyes that interact non-covalently with hydrophobic regions of proteins gave much more consistent results, and we eventually settled on a merocyanine dye, known by the trade name of NanoOrange (Molecular Probes, Invitrogen), that produces a large increase in fluorescence quantum yield upon interaction with detergent-coated proteins [30
]. It stood out among others for its high sensitivity, wide dynamic range, insensitivity to nucleic acids, relatively equivalent staining of most proteins, and above all, ease of use and high reproducibility. It is compatible with commonly used paraformaldehyde fixation, allowing the maintenance of morphological features for high resolution imaging studies. Furthermore, it can be combined with antibody-based immunofluorescent detection of specific proteins for multiparameter assays, thus providing both specific target and total protein information in the same specimen. In this communication we present optimized protocols for these new staining methods.
Using NanoOrange staining we confirmed, using single-cell image analysis, the increase of protein in senescent HDF, and extended these observations to MTF, providing additional generality to these findings. We substantiated the increase of protein in the nuclear compartment, which has not been generally noted or investigated. We were initially interested in the nucleus for practical reasons, since it can be easily and precisely identified by DAPI staining, which also provides a reliable way to normalize the protein signals. These are important quality control considerations when working with tissues. Interestingly, our data indicate, by the increase of the ratio of mean protein intensity to mean DNA intensity, that senescent cell nuclei are more dense in protein.
Multiparameter studies combining NanoOrange staining with immunofluorescence clearly showed that the increase in nuclear protein content is not caused by a general increase in the abundance of all nuclear proteins. Neither LMNA, a major structural protein, or histones H3 and H2B, were increased. As previously documented [11
] the heterochromatin component histone mH2A increased in senescent cells, but because of its relatively low abundance is unlikely to account for a significant fraction of the increase in total protein. The nuclear protein (or proteins) responsible for this phenomenon thus remain to be identified.
In mouse liver we observed, in addition to the increase in signal, a localization of NanoOrange staining into prominent foci. These foci colocalized to a large extent with DAPI foci, which are known to correspond to centromeric heterochromatin. It thus appears that in mouse liver centromeric heterochromatin becomes more protein dense with age. The extent to which this accounts for the overall increase in signal is currently under investigation. It is interesting to note that while NanoOrange foci are also often seen in HDF and MTF, they are not exclusive to senescent cells, and are unlikely to be Senescence Associated Heterochromatin Foci (SAHF) since they do not colocalize well with either DAPI or mH2A foci. It is thus possible that the increase in total protein in vitro and in vivo may proceed by different mechanisms.
The combination of NanoOrange staining with automated image analysis platforms is highly desirable. First, the inherent biological variability of the specimens makes it essential to generate a large number of observations (ideally, thousands) to produce data that are highly statistically significant. The CellProfiler software [40
] allowed us to achieve this in a reasonable time frame. Second, CellProfiler can be trained to recognize a variety of subcellular structures, for example, we are developing a pipeline to score liver cells for the age-associated NanoOrange nuclear foci. Third, CellProfiler can score a large number of parameters such as granularity, texture, etc., that may not have immediate biological interpretations, but should provide useful starting points for the investigation of age-associated changes in different tissues. In this context it should be noted that the cytoplasm, which we have not explored in this preliminary report for practical reasons, is very “feature rich”, and it is likely that novel information can be mined in this area. Finally, the CellProfiler Analyst platform [41
] enables interactive, machine learning-based data explorations to define and score complex and novel visual phenotypes, and even use these to distinguish small subsets of cells with specific characteristics. We believe these methods will be of wide interest to biologists in many fields in addition to aging.