Machine-based method for multiplex in situ molecular characterization of tissues by immunofluorescence detection

Immunofluorescent staining is an informative tool that is widely used in basic research. Automation of immunostaining improves reproducibility and quality of the results. Up to now, use of automation in immunofluorescent staining was mostly limited to one marker. Here we present tyramide signal amplification based method of multiple marker immunofluorescent detection, including detection of antibodies, raised in the same species, in tissue sections and cultured cells. This method can be beneficial for both basic and clinical research.

target expression in the tissue. By combining this characteristic of the TSA protocol with the consistency of automated staining results, we are able to reproducibly perform successful IF experiments.
Characterizing co-expression and co-localization of multiple antigens is an important and often-used methodology in research as well as in clinical settings. However, reliable multiple-marker IF staining can be difficult to achieve. The specificity of each antibody must be validated in single staining using proper controls and must be retained when multiple antibodies are applied. Generally, antibodies raised in different species are used to prevent cross-reactivity. However, it is not always possible to find optimal antibodies of interest made in different species. Even if such antibodies are identified, achieving successful detections is not guaranteed.
Methods for double staining with antibodies derived from the same species have been published: 1) Adjacent thin sections are stained separately and images are superimposed 10,11 . 2) Primary antibodies of a particular isotype are detected with secondary antibodies, specific only for that specific isotype 12 . 3) Primary antibodies are directly conjugated to fluorophores, enzymes or haptens [13][14][15] . 4) Saturation of epitopes in double IF or IHC to prevent non-specific binding of antibodies of the same species was done with fluor-or respectively HRP-or Alkaline phosphatase-conjugated IgG (Fab) used as secondary antibodies [16][17][18] . 5) Tyramide amplification combined with conventional detection was used in double IF with frozen skin sections 19 . Unfortunately, all of these methods, except TSAbased, either have limitations or simply do not work under automated staining conditions.
In our laboratory we discovered that combining consecutive staining method with TSA-based detection allows use of multiple antibodies, including antibodies raised in the same species, in automated staining. Here, we present examples from a small subset of multichannel and same-species fluorescent stainings of paraformaldehyde and formalin fixed paraffin-embedded sections of human and mouse tissues. To our knowledge, this is the first published demonstration of successful and reliable machine-based automated IF detection of multiple antigens.

Methods
Adult mouse tissues (liver, testis, ovary and spleen) and E13.5 mouse embryos were fixed with freshly made 4% paraformaldehyde (PFA) in PBS (Sigma Aldrich) overnight at 4uC, processed with Leica ASP6025 tissue processor (Leica Microsystems), embedded in paraffin and sectioned at 5 mm. Human tumors were donated by the Pathology Department without any unique patient identifiers except diagnosis. Human tissues were fixed in 10% Neutral Buffered Formalin (NBF), unless otherwise indicated, processed with Leica ASP6025 tissue processor (Leica Microsystems), embedded in paraffin and sectioned at 5 mm.
All experiments were performed in accordance with approved guidelines established by MSKCC IACUC and IRB. The protocol #01-11-026 for mouse experiments was approved by MSKCC IACUC. Human tissues were used according to the research exempt from IRB/PB review SOP# FO-303 (Part 3.4).
The IF staining was performed using Discovery XT processors (Ventana Medical Systems). Sections were deparaffinized, conditioned and the antigens were retrieved with proprietary buffers, EZPrep and CC1 (Ventana Medical Systems). Slides were blocked for 30 minutes with peptide-based blocking reagent Background Buster (Innovex). Primary antibodies were applied at optimized concentrations previously determined on control tissues (see Table 1). Sections were incubated with primary antibody, followed by 60 minutes incubation with biotinylated secondary antibodies (Vector Laboratories) against immunoglobulins of the primary antibody source species at 7.5 mg/ml in PBS with 2% BSA and 1.5% normal serum of secondary antibody source species. The detection was performed with Streptavidin-HRP D (DABMap kit, Ventana Medical Systems), followed by incubation with one of the following Tyramide Alexa Fluors (Invitrogen): 488 (cat# T20922), 546 (cat# T20933), 568 (cat# T20914), 594 (cat# T20935) or 647 (cat# T20936) prepared according to manufacturer instruction with predetermined dilutions. Slides were counterstained with DAPI (Sigma Aldrich, cat# D9542) at 5 mg/ml in PBS (Sigma Aldrich) for 10 minutes, mounted with Mowiol (Calbiochem) and kept at 220uC.
The stainings were performed consecutively. The detection for each marker was completed before application of next antibody. The best sequence of antibodies for multiple staining was determined for each combination. When multiple staining was done with primary antibodies raised in the same species, tissue sections were ''saturated'' by re-incubation with the primary antibody they were just stained with, for 1 hour at room temperature in the dark, followed by washing in reaction buffer (Ventana Medical Systems). The experiment continued with the next primary antibody. In some cases slides were fixed in 4% PFA in PBS for 10 minutes between the stainings to stabilize the IgG complex of the first staining and prevent non-specific binding.

Results
We established a reliable and reproducible method of automated multiple marker IF detection on the Ventana Discovery XT instrument. The method employs the use of biotinylated secondary antibodies, Streptavidin-HRP conjugate and tyramide-fluorophore conjugate to detect and amplify the signal. Introduction of this method allows the sequential detection of multiple markers even when primary antibodies are raised in the same species. Earlier experiments, with other standard methods of fluorescent detection, produced suboptimal or even negative results. Only TSA-based system provided reliable and reproducible results. In Figure 1, we present the comparison of three detection methods: biotinylated secondary antibodies plus Streptavidin-HRP followed by TSA-Alexa 488 (Panel A); secondary antibodies conjugated with Alexa 488 (Panel B); and biotinylated secondary antibodies plus Streptavidin-Alexa 488 (Panel C), used to stain CD31 in mouse 13.5 embryo. One can appreciate that in cases of strong CD31 staining, detectable signal is present in all detection methods (although at very different levels), whereas in smaller blood vessels with lower expression levels of CD31, the signal is only detected with TSA-based method. We have numerous similar results for manual and automated stainings of both frozen and paraffin embedded tissues. We applied this method to a wide range of tissues from different species (mouse, human, rat, rabbit, fish, Drosophila). In this article we present only a few examples of successful multiple stainings. Spleen is one of the tissues that is rather complicated for staining, as it is part of the immune system and contains many types of cells involved in the immune response. We used several combinations of markers to characterize the immune response in spleen.
In Figure 2, we present the staining of human spleen sections with CD4 antibody (mouse) in green, CD3 antibody (rabbit) in red, CD8 antibody (mouse) in white and DAPI in blue. CD4 and CD8 antibodies were used to distinguish the two major subpopulations of T cells that are CD3-positive. DAPI was used to stain cell nuclei. One can appreciate that the markers for two subpopulations -CD4 and CD8 -are distinct and appear to be expressed on different cells (Panel A1), whereas each subpopulation-specific marker is coexpressed with CD3 (Panel A2 for CD4/CD3 and A3 for CD8/CD3). Sequential IF staining with immune markers were also successfully performed using mouse spleen sections. However, when mouse/rat antibodies were used to stain mouse spleen, in addition to the specific signals, nonspecifically labeled cells were frequently present ( Figure 3). We consider those cells plasmacytes, rich with immunoglobulins, thus always stained with the secondary anti-mouse/rat antibodies. The same cells were also detected when isotype control IgGs were used instead of primary antibodies or in case when no primary antibodies were used ( Figure 3).
In another example, we present data from a successfully performed quadruple immuno-staining using the TSA method, shown in Figure 4. Human kidney tumor is stained for CD31 (mouse antibody) in green, Vimentin (guinea pig antibody) in red, E-Cadherin (mouse antibody) in white, and PCNA (mouse antibody) in magenta plus DAPI in blue. A representative fragment of the tissue section shows that none of these markers are co-localized and the signals are strong. Panels B, C and D reveal staining details of several regions of the tissue in higher magnification. As in any sequential antibody staining protocol, the sequence of antibodies used can greatly affect the quality of specific staining and levels of background. Importantly, the appropriate sequence could be specific for any combination of antibodies and must be determined by trial and error.
We have many examples of successful staining of different tissues combining antibodies from the same or different species and stained in different sequences (Supplemental Table 1). Figure 5 includes examples of multiple staining of mouse tissues with the antibodies derived from the same species. Panel A shows mouse ovary stained with three rabbit antibodies -VASA, Ki67 and Laminin. Panel B shows mouse liver stained with two mouse antibodies -E-Cadherin and N-Cadherin. Each of the combinations was tested several times and each staining included single marker controls as well as nonimmune isotype controls. In addition, each marker was previously tested and validated by immunohistochemical staining with DAB detection.
Human tissues presumably have higher rate of specificity, as the antibodies against human antigens are more vigorously tested and more readily available. However, more frequently, we stain embryonic and adult mouse tissues that pose greater risk of non-specific binding of antibodies. Figure 6 shows examples of successful multicolored stainings of mouse E13.5 embryo. Panel A represents an overview of a paraffin section through a E13.5 embryo, as well as detailed views of staining with four markers -CD31 (rat antibody), a blood endothelial cell marker in green; Lyve 1 (goat antibody), for Paraformaldehyde fixed, paraffin embedded 5 mm sections of E13.5 mouse embryo were stained with CD31 primary antibody (see Methods for details) and nuclei were stained with DAPI. Panel A shows detection with biotinylated goat anti-rat secondary antibodies followed by Streptavidin-HRP and TSA-Alexa488 amplification. Panel B shows detection with goat anti-rat secondary antibodies directly conjugated with Alexa488. Panel C shows detection with biotinylated goat anti-rat antibodies followed by Streptavidin-Alexa488. Panel D shows staining with isotype control IgG detected with biotinylated goat anti-rat antibodies, followed by Streptavidin-HRP and TSA-Alexa488. The inset in each panel represents a closer view of the indicated region, with CD31 signal showing in grayscale. White arrows point to autofluorescence emitted by the eurethrocytes, which persists in all staining methods. In sum, our data show how powerful automated immunofluorescence detections can be and how much information it can bring to both clinical and basic research.

Discussion
In this report, for the first time we present the method of automated IF staining with multiple markers in the same tissue sections. Furthermore, our protocol allows researchers to use antibodies raised in the same species. To prevent non-specific binding of the second primary antibodies to unbound epitopes of secondary antibodies, we perform ''saturation'' with primary antibodies, used in the first staining (normal IgGs can also be used). We utilize the TSA detection system that involves biotinylated secondary antibodies, subsequently labeled with Streptavidin-HRP and amplified with tyramide-fluorophore conjugates. When IHC staining is performed manually, this method is the best for detection of low-expression targets [19][20][21][22] . For an automated immunofluorescence protocol, TSA is the only method that produces reliable results (see Figure 1). In our hands, all other methods have resulted in either low signal, or no staining at all (see Figure 1). It is essential to note that the concentrations and incubation times of primary antibodies used for the IF stainings are the same as we use for chromogen-based IHC stainings. It is also important, that TSA-based method of IF detection, as opposed to TSA amplification in IHC detection, allows relative quantification of the signal levels.
Although in our laboratory we use Ventana Discovery machines, this approach could be adopted for any open automated platform that can utilize HRP-based detection. In a testing of a Leica Bond RX machine, we were able to perform IF staining of paraffin sections with CD31, isotype control and no primary antibody control. Panel A shows the staining of blood vessels with CD31 along with non-specifically stained cells (red), Panel B represents just the non-specific staining (red) with isotype control (rat IgG) and in Panel C same ''plasmacytes'' are observed in staining that skips primary antibodies or isotype control. using open protocols and the same reagents we used for staining on the Ventana machine.
While one may take advantage of the reproducibility and convenience of automated staining, obstacles that befall any immuno-staining -manual or automated -are still present. For example, some antibodies are simply incompatible with one another: certain combinations of antibodies will unavoidably give non-specific staining or cross-reactivity. It is possible that the antibodies recognize similar fragments of antigens, or some tissue components are ''sticky'' and attract antibodies non-specifically. Automation and signal amplification with tyramide does not alleviate all problems. However, it greatly increases the frequency of successful detections.
As in manual staining, the sequence of antibodies can significantly increase or decrease the level of specific signals and non-specific background. This can be explained by the differences in the stability of the epitopes, the conformation of the target proteins, the effect of antigens' proximity and/or other less clear factors related to the execution of experiments with automated machines. Researchers must still determine the optimal sequence of antibodies through trial and error.
The choice of secondary antibodies should also be adjusted to eliminate possible non-specific binding. For example, widely used goat anti-rabbit secondary antibodies would recognize rabbit antimouse antibodies and produce non-specific staining in multiplex IF experiments that include primary antibodies produced in goat. Again, this is an important aspect of any multi-antigen staining and often is neglected as the major attention is focused on the primary antibody specificity and clonality.
The staining approach we describe in this article allows researchers to utilize both paraffin sections and cryosections. Our experience suggests that IF staining could be carried with paraffin sections from clinical specimens, even archived. That widens opportunities for assessment of a large number of tissues with well-preserved morphology, including clinically important human tissue specimens. While we presented here IF staining of formalin fixed paraffin-embedded tissues, using the same protocols we have successfully stained frozen tissues, as well as fixed cells.
Another key aspect of our method is that the stainings are performed consecutively. That allows analysis of the results from each staining step to be performed sequentially. The most appropriate marker for the following staining could thus be identified. Importantly, slides can be digitally scanned and the data stored at any step of the sequential staining. As we demonstrate, the presented method of automated IF staining is a powerful and efficient tool that enables reproducible, reliable, high quality staining data to be obtained. Additionally, IF signals can be accurately and reliably quantitated and correlated to antigen expression levels. Such results can bring immeasurable contribution to basic and clinical research.   Localization of multiple markers in the same tissue section provides unique insight about spatial, cell -type and even organelle -type specific distribution of molecules of interest. It allows deeper understanding of the tissue during development, in normal adult state, in disease and after treatment.