Improved Multiplex Immunohistochemistry for Immune Microenvironment Evaluation of Mouse Formalin-Fixed, Paraffin-Embedded Tissues

Using immunohistochemistry (IHC) to evaluate immune cell dynamics in vivo complements immunophenotyping via flow cytometry (1, 2), providing a method for validation. Further, with the advent of immunotherapy to treat cancer and other diseases, it has become increasingly important to evaluate infiltrating immune cells in the context of the whole tissue (3–5). Determining the tissue distribution of specific Ags, especially immune markers, is necessary for the diagnosis and investigation of disease. In cancer, this can assist in the understanding of immune cell behavior in the tumor microenvironment at primary and metastatic sites (6, 7).

Although multiple groups have made strides regarding multiplex IHC of human tissues for immune cell characterization (8–10), protocols using mouse formalin-fixed, paraffin-embedded (FFPE) tissues are still underdeveloped. Many epitopes are formalin fixation–sensitive and not easily detected. Historically, detection of formalin-sensitive epitopes has relied on frozen tissue (11, 12), which comes with the disadvantage of cumbersome storage and compromised tissue morphology.

Zinc salt fixation has been proposed as an alternative to formalin fixation as a strategy to retain tissue morphology while resulting in less epitope masking (13, 14). Zinc salt fixation may indeed have advantages; however, there are only limited IHC studies comparing the two fixation methods. Additionally, there are concerns that zinc-based fixatives may not penetrate tissues as effectively as formalin (15). Further, to the best of our knowledge, the long-term stability of zinc-fixed tissue is unreported. Finally, formalin fixation remains the most common fixation method for mouse tissues.

In this study, we demonstrate multiplex IHC of mouse immune cell markers with an improved Ag retrieval protocol that works efficiently for most Abs. We used specific markers to detect populations of B cells, subtypes of T cells, NK cells, macrophages, dendritic cells (DCs), granulocytes, and total myeloid cells. We assessed Ab specificity by using tissue from wild-type, SCID, NOD SCID γ (NSG), and control- or clodronate liposome–treated mice. Finally, we demonstrate the utility of the protocol in evaluating the immune landscape in a preclinical model of breast cancer.

All materials are available from commercial sources or can be derived using methods described in this study. All relevant data are reported in the article.

Male and female BALB/c (BALB/cfC3H strain) and SCID mice were purchased from the on-campus University of Texas (UT) Southwestern Mouse Breeding Core (Dallas, TX). Tissue from NSG mice was a kind gift from Dr. S. Morrison (UT Southwestern Medical Center). All animals were housed in a pathogen-free environment with access to food and water ad libitum.

Control (“Encapsomes”) and clodronate liposomes (“Clodrosomes”) were obtained from Encapsula NanoSciences (Brentwood, TN). Eight-week-old BALB/c male mice were injected i.p. with 300 μl of Encapsomes (control) or Clodrosomes. Mice were euthanized 72 h postinjection, and tissues were fixed in 10% neutral-buffered formalin for 48 h before embedding in paraffin.

The 4T1 breast cancer cells were obtained from the American Type Culture Collection (Manassas, VA). The cells were cultured in DMEM (Sigma-Aldrich) supplemented with 10% FCS (Rocky Mountain Biologicals) at 37°C in a humid atmosphere with 5% CO₂.

The 4T1 cells (10⁵) were injected orthotopically into the inguinal mammary fat pat of 8-wk-old female BALB/c mice. Mice were euthanized 3 wk postinjection. The established tumor was resected and fixed in 10% neutral-buffered formalin for 48 h before embedding in paraffin. All experiments were approved by the Institutional Animal Care and Use Committee at UT Southwestern and conducted in accordance with an approved protocol.

All tissues were fixed in 10% neutral buffered formalin for 48 h on a shaker at room temperature. Tissues were washed and stored in PBS at 4°C until embedded. Paraffin embedding was performed by the UT Southwestern Molecular Pathology Core.

Before sectioning, FFPE tissue blocks were incubated in 70% ethanol with 10% glycerol for 5 min to soften, followed by a quick rinse in distilled water. Then, 5-μm sections were cut using a Leitz 1512 Rotary Microtome and placed onto a water bath at 40°C for 2 min to reduce wrinkles. Sections were transferred to positively charged slides (no. SFH1103; Biocare Medical) and allowed to dry overnight before staining.

Slides were warmed for 10 min in a 60°C oven before following the deparaffinization protocol. The deparaffinization and rehydration protocol entails incubation in xylene (3× for 5 min), 100% ethanol (2× for 2 min), 95% ethanol (2× for 2 min), 70% ethanol (2× for 2 min), 50% ethanol (2× for 2 min), and PBS (1× 3 min). To prevent sections from folding or falling off the slides during Ag retrieval and Ab stripping, the deparaffinized slides were fixed in 10% neutral buffered formalin for 30 min, followed by incubation in PBS (2× for 3 min).

For Ag retrieval, 250 ml of Ag retrieval buffer (10 mM Tris-HCl and 1 mM EDTA with 10% glycerol [pH 9]) was preheated to 80°C in a bucket inside a pressure cooker (Biocare Medical) filled with 500 ml of water. Slides were then placed inside the buffer and heated at 110°C for 18 min (∼4–5ψ). Slides were allowed to cool at room temperature for 30 min, followed by a PBS rinse. Tissue sections were surrounded by a hydrophobic barrier using a PAP pen and blocked for 10 min with 2.5% goat serum and 2.5% horse serum (Vector Laboratories) or for 30 min with Rodent Block M (Biocare Medical) if the secondary host was goat, horse, or rat, respectively.

Incubation with primary Ab was performed on a shaker for 1 h at room temperature or at 4°C overnight. Slides were washed in PBS with Tween (PBSt; 0.05% Tween 20 plus 2 mM EDTA; 3× for 5 min) before adding the “polymer” HRP- or alkaline phosphatase (AP)-conjugated secondary Abs (ImmPRESS; Vector Laboratories). Sections were incubated with secondary Ab for 30 min on a shaker, followed by one PBSt (0.2% Tween 20 plus 2 mM EDTA) and two PBSt (0.05% Tween 20 plus 2 mM EDTA) washes, 5 min each. Sections were then incubated with chromogenic or tyramide signal amplification (TSA) substrates. For multiplex staining, Ab stripping was performed in 10 mM citrate buffer (pH 6.2) plus 10% glycerol in a pressure cooker at 110°C for 2 min before probing with the next primary Ab.

For chromogenic detection, Betazoid DAB, Warp Red, and Ferangi Blue (BDB2004L, WR806S, and FB813H, respectively; Biocare Medical) were used. Substrate incubation time was determined by monitoring signal development using a microscope. After developing with chromogenic substrates, slides were counterstained with hematoxylin. Bluing was performed by running tap water over the slides for 30 s. Slides were coverslipped using VectaMount (H-5501; Vector Laboratories) and scanned at 20× using the Hamamatsu NanoZoomer 2.0-HT.

For fluorescence detection, we used a TSA detection system (PerkinElmer). The following fluorophores were used: OPAL 520, OPAL 570, and OPAL 690. DAPI was used for nuclear counterstaining. To reduce the autofluorescence, slides were incubated at the end in TrueBlack Lipofuscin Autofluorescence Quencher (no. 23007; Biotium) diluted 20-fold in 70% ethanol for 30 s. Slides were coverslipped using ProLong Gold (no. P36931; Life Technologies). Slides were scanned at 20× using the Zeiss Axio Scan.Z1 (Whole Brain Microscopy Facility, UT Southwestern). DAPI, AF488 (for OPAL 520), AF555 (for OPAL 570), and AF660 (for OPAL 690) channels were used to acquire images. The exposure time for image acquisition was between 5 and 100 ms.

Primary Abs were diluted in PBSt (0.05%) plus 2 mM EDTA plus 0.5% serum (goat or horse, based on the host of the secondary Ab), or Renaissance Background Reducing Diluent (PD905; Biocare Medical) when incubating sections with more than one primary Ab simultaneously. The primary Abs and dilution factors used are displayed in Table I. Polymer secondary Abs were purchased from Vector Laboratories (ImmPRESS HRP anti-rabbit IgG [no. MP-7401], HRP anti-rat IgG [no. MP-7404], AP anti-rat IgG [no. MP-5404], and AP anti-rabbit IgG [no. MP-5401]).

Single-cell suspensions were prepared from C57BL/6 mouse spleens. The 1 × 10⁶ cells were stained with CD11b PE, Ly6C PerCP-Cy5.5, Ly6G PE-Cy7 (eBioscience), F4/80 APC (Bio-Rad AbD Serotec Limited), and FcγR Abs as previously described (16, 17). Data were acquired on the FACSCanto II (BD Biosciences) and analyzed with FCS Express (De Novo Software). FcγR expression by macrophages (F4/80⁺), monocytes (CD11b⁺Ly6C⁺), and neutrophils (CD11b⁺Ly6G⁺) is demonstrated by mean fluorescence intensity.

Image analysis was conducted using Fiji software (18). During the analysis, eight-bit images were thresholded at the same intensity value for a specific staining (e.g., CD3, CD11c, etc.) and converted to binary format. To calculate tissue areas that are double-stained or have specific staining patterns (e.g., CD8⁺ and CD4⁻), an “Image Calculator Function” was used. Then, the area fraction of the signal in resulting binary images was calculated. All calculated area fractions of different staining were normalized using DAPI.

We sought to develop an improved protocol for IHC on mouse FFPE sections with the following requirements: 1) reproducibility of staining across different tissues from different experiments; 2) consistent results with most Abs used without laborious Ab-specific Ag retrieval optimization; and 3) ability to perform multiplex IHC.

In accord with previous reports, we found that Tris/EDTA (pH 9) Ag retrieval buffer was optimal for most primary Abs (19). Further, we found that the addition of 10% glycerol significantly enhanced Ag retrieval in multiple cases (Supplemental Fig. 1A–L). Additionally, we found that adding a short 10% neutral buffered formalin fixation step after deparaffinization prevented tissue folding, which can be an issue with higher pH buffers during heat-induced epitope retrieval (HIER). This did not impact epitope detection (data not shown).

For Ab stripping, some groups use a short HIER-like step in citrate buffer (pH 6.2), whereas others have reported success with incubating the slides in a 100-mM glycine/Tween 20 buffer (pH 10) at room temperature (14). We found that the glycine buffer–based stripping method failed to remove the primary Ab (Supplemental Fig. 1N, 1Q). In contrast, the short HIER-like step in citrate buffer worked effectively (Supplemental Fig. 1O, 1R). We suspect heat denaturing of high-affinity Abs is particularly important in this process.

To demonstrate consistency across detection methods, we first performed single staining on spleens using three different Abs, starting with fluorescence (TSA) detection. Following a stripping step, we reprobed the same tissue with the same Abs and used chromagen detection (Supplemental Fig. 2). Cells were stained identically with each method.

To evaluate the effect of the stripping step on Ag retrieval, we stained spleens after the primary Ag retrieval step (“Round 1,” Supplemental Fig. 3). Second, we stained different spleen sections for the same markers after performing Ag retrieval and a single stripping step (“Round 2,” Supplemental Fig. 3). Third, we repeated this on spleens that were subjected to Ag retrieval and two subsequent stripping steps (“Round 3,” Supplemental Fig. 3). For the three Abs tested on spleens (CD3, CD19, and Pax5), we did not observe significant differences in detection before or after the tissue was exposed to multiple stripping steps. However, this is not the case for every Ab/epitope (e.g., F4/80 [D2S9R]; Supplemental Fig. 3K, 3L). Thus, laboratories should perform such testing to determine the order of staining for multiplex protocols involving Ab stripping steps.

After the primary Ag retrieval step and multiple iterations of stripping in multiplex staining, we found that the nuclear signal, via DAPI or hematoxylin, is lost if the staining is done over a longer timeframe (hours versus days; data not shown). We suspect that nuclear DNA is vulnerable to environmental DNase contamination after extensive destruction of formalin cross-links. This issue was resolved by adding EDTA to wash buffers and Ab diluents.

Applying our revised protocol, we tested over 35 primary Abs specific for mouse immune subsets and developed a panel of lymphoid and myeloid cell markers (Table I). These markers were chosen based on ImmGen expression data (Fig. 1) (20), previous reports (16, 17, 21–24), and Ab availability. Using chromagen and fluorescence (via TSA) detection methods, we observed that each marker produced expected staining patterns corresponding with known localization and expression by the respective cell types of interest (Supplemental Fig. 4) (25).

To validate Ab specificity, we used two strategies. First, we stained tissue in which cell types/markers of interest are expected to be decreased compared to control tissue. For the lymphoid panel, we stained spleens from wild-type, SCID, and NSG mice (Fig. 2A–G1). Compared to wild-type, we observed a decrease in lymphoid cell markers in SCID and NSG spleens. For example, CD3 and CD19 (T cell and B cell markers, respectively) are completely absent in SCID and NSG tissue. As a positive control, CD11b, mostly a myeloid-lineage marker, was consistent across all groups (Fig. 2B1–D1).

To test macrophage markers, we stained livers from control- or clodronate liposome–treated mice, as clodronate liposomes deplete professional phagocytic cells, primarily macrophages (26, 27). Macrophage staining was markedly reduced in clodronate liposome–treated mice (Fig. 2H1–Q1).

For our second strategy to confirm Ab specificity, we used multiple Abs for the same marker or cell type when possible. For example, we used two different Abs for F4/80 (Fig. 2H1–K1). For B cells, we used two different B cell–specific markers, CD19 and Pax5 (Fig. 2J–O).

When developing the Ab panels, we found that for some cell types (e.g., DCs, neutrophils, and NK cells), Abs for cell-specific markers were not available. To address this, we used multimarker staining strategies.

For example, we developed a strategy to stain NK cells using CD3 and Zap70. CD3 is expressed by T cells, but not NK cells (Fig. 1) (20). Zap70 is expressed by T cells and NK cells (Fig. 1) (20). Because SCID and NSG mice lack T cells, CD3 staining is absent in SCID and NSG spleens compared with wild-type (Fig. 2A–C). In contrast, some Zap70⁺ cells remain, presumably NK cells (Fig. 2D–F). Combining these markers, we stained spleens for CD3 and Zap70 using both chromagen and fluorescence detection methods, demonstrating the presence of CD3⁺Zap70⁺ and CD3⁻Zap70⁺ cells, corresponding to T cells and NK cells, respectively (Fig. 3A, 3B). In the case of chromagen staining, we first stained for CD3 using a dark brown chromagen (diaminobenzidine), intentionally obscuring further staining wherever the chromagen is deposited. We subsequently performed Zap70 staining using a red chromagen (Warp Red). Through this strategy, it is easy to identify Zap70 single positive cells using chromagen detection (Fig. 2A, inset).

We also developed a strategy to detect DCs. Although CD11c has long been used as a DC marker, it is not cell type exclusive, as it is may be expressed by macrophage subsets and other immune cells (Fig. 1) (20). MHC class II (MHCII; H2-Ab1) is also not exclusive, but it is expressed more consistently across DC subsets and is found on fewer cell types (Fig. 1) (20). Because MHCII is expressed by splenic B cells and macrophages, we first stained for B cell and macrophage markers, Pax5 and CD163, respectively. By combining CD163 and Pax5 signal with the same chromogen or fluorophore, we demonstrate the effective use of a “dump channel” applied in an IHC setting (Fig. 3C, 3D). Staining for MHCII second with a different chromagen or fluorophore revealed a distinct population of MHCII⁺Pax5⁻CD163⁻ cells, which we believe to be DCs (Fig. 3C, 3D).

Next, we developed a strategy to distinguish neutrophils from macrophages. Using knowledge from the literature and considering the available Abs for mouse targets, we stained for CD163 and FcgR4. According to ImmGen (Fig. 1) (20) and previous reports (16, 17), FcgR4 is expressed by macrophages and neutrophils. This is corroborated by flow cytometric analysis (Fig. 3G). Based on this information, we expect that splenic macrophages are CD163⁺FcgR4⁺ and neutrophils are CD163⁻FcgR4⁺. Performing this multiplex staining reveals cells that follow these staining patterns (Fig. 3H, 3I).

Finally, we tested our immune panels and multiplex staining on 4T1 tumors, which is an immune cell–heavy model of breast cancer. We did multiplex IHC staining using chromogen and fluorescence detection methods (Fig. 4). Fig. 4A and 4B shows CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells. We also detected CD3⁻CD4⁺ and CD3⁻CD8⁺ cells, which may be NK cells or DCs. Notably, all CD3⁺ T cells were labeled with either CD4 or CD8, demonstrating the specificity and sensitivity of the staining, as the presence of double-negative or double-positive T lineage cells is unreported in this model.

Fig. 4D (inset, right arrow) shows the presence of CD11c⁺ macrophages. This corresponds to tumor-associated macrophages, as resident tissue macrophages in the mammary fat pad are CD11c⁻ (28). We also detected F4/80⁻CD163⁻CD11c⁺MHCII⁺ DCs (Fig. 4D, inset, left arrow).

Fig. 4G and 4H quantitation of different immune cell types in 4T1 tumors normalized to the DAPI signal. The 4T1 tumors have been reported to have high neutrophil infiltration (29). Consistent with this, our analysis shows a high frequency of F4/80⁻CD163⁻FcgR4⁺ cells, which are presumably neutrophils (Fig. 4E, 4F).

Comparatively, we also found modest numbers of tissue macrophages (F4/80⁺CD163⁺CD11c⁻), DCs (F4/80⁻CD163⁻CD11c⁺MHCII⁺), and T cells (CD3⁺CD4⁺ and CD3⁺CD8⁺), data that are consistent with the findings of other groups (30–32). In sum, these data illustrate how an effective multiplex IHC strategy can be exploited to investigate the immune dynamics of the tumor microenvironment.

In this study, we produced a modified FFPE IHC protocol for mouse tissue that provides a more effective Ag retrieval method. Ag retrieval has been a longstanding challenge in the field, often requiring different retrieval protocols for different epitopes (14, 33, 34). Our protocol works consistently well for all of the Abs tested. The use of glycerol in the Tris/EDTA Ag retrieval buffer greatly increases the detection of multiple Ags and may be a better standard for mouse FFPE tissues. Also, the use of EDTA in the wash buffers prevents degradation of DNA by nucleases, preserving nuclear morphology, a consideration that is absent in other multiplex protocols. Finally, we found that treatment with 10% formalin immediately prior to Ag retrieval helps to maintain tissue integrity and reduces loss of tissue during multiple rounds of stripping and reprobing.

We also propose a novel strategy of using multiple markers to stain for cell types for which unique markers/Abs have not been identified or are not available commercially. This strategy can be particularly helpful to identify DCs, neutrophils, and NK cells in the tumor microenvironment. With new emerging immune regulators (35) and therapies, this strategy provides an excellent tool to investigate the effect of drug treatment and prognosis of disease.

Overall, our protocol of multiplex IHC staining on mouse tissue provides an improved strategy for reproducible staining. Using our strategy, and by properly planning the sequence of target Abs, visualization of multiple epitopes is only limited by the number of filter channels in the microscope. This protocol has the potential to improve the utility of multiplex IHC to investigate the immune system in vivo, including the analysis of the tumor immune microenvironment and mechanisms of action of immune-related drugs in preclinical models.

We thank D. Primm for editorial assistance and Drs. E. Koay and A. Zijlstra and members of the Brekken Laboratory for advice and helpful discussion.

The authors have no financial conflicts of interest.