Systematic High-Content Proteomic Analysis Reveals Substantial Immunologic Changes in Colorectal Cancer

Introduction

Colorectal cancer is a common malignancy with an incidence of nearly 1 million cases per year worldwide. Although tremendous efforts in the therapy of colorectal cancer have improved treatment and survival, the 5-year survival rate remains ∼50%, and the annual mortality of 492,000 per year substantiates the need for new or complementary therapeutic strategies ( 1).

Given the immunogenetic properties of colorectal cancer, promising results with immunotherapy have been published, restoring either the disease-free and overall survival with immunization or restoring the immune response ( 2). However, multifactorial changes to the host defense caused by colorectal cancer such as a loss of human lymphocyte antigen (HLA) class I processing or signaling molecules, escaping of death receptors, or impaired natural killer (NK) cell function are less understood and contribute to tumor escape from immune recognition and, thus, poor prognosis.

Because an exact linear association between genome, transcriptome, and proteome of a cell does not exist, and translation and posttranslational modifications of proteins are not apparent from the DNA or mRNA sequence, providing a contribution to better understand the host response to colorectal cancer requires analyzing expression, function, and regulation of the entire set of proteins encoded by an organism – the aims of the emerging field of proteomics ( 3– 5). This information will be invaluable for understanding how complex biological processes occur at a molecular level, how they differ in various cell types, and how they are altered in disease states. To analyze these complex biological processes in situ, a highly flexible multiplex detection system, the so-termed Multi-Epitope Ligand Cartography (MELC)-technology, was applied. This novel technology allows the simultaneously visualization of dozens of proteins in a structurally intact single cell as well as in the complexity of tissues. Advanced statistical data analysis and visualization software was used to process and analyze the highly multipart information generated by MELC. With these unique features, the possibility to correlate cellular localization of proteins with their function and consequently exploring highly specific protein networks is given. This multidimensional microscopic robot technology allows not only high-throughput protein recognition but, furthermore, the combination of localized proteins, providing the opportunity to generate a protein colocalization map. Recently, the identification of disease-relevant protein networks arising from detection of such protein expression patterns in inflammatory skin and bowel disease by using the advantages of the MELC technology has been shown ( 6, 7).

In our study, we performed for the first time a systematic analysis of key immune function–related proteins and in situ detection of their modification in the tissue of colorectal cancer patients. By comparing the mucosal toponomic picture in colorectal cancer with that of ulcerative colitis, we hope to uncover not only changes in the expression and distribution of proteins critically involved in the pathogenesis of colorectal cancer but also explore differences in the host response in malignancy and inflammation.

Materials and Methods

Patients and Sample Preparation

A total of 10 biopsies per group (healthy controls, inflamed tissue of ulcerative colitis, and colorectal adenocarcinoma) were analyzed. All biopsies were derived from the colonic mucosa and were obtained endoscopically. The macroscopic aspect of normal, inflamed, or neoplastic colonic mucosa was verified and confirmed by the histopathologic result. The median age of the colorectal cancer patients was 64 years (range, 37–81 years), 6 of them were males and 4 were females. The main histopathologic characteristics were low differentiation (6/10 G3) and deep mucosal invasion (7/10 pT3/pT4). All adenocarcinomas were located in the colon, not in the rectum. In all cases, the biopsies were taken at time of initial diagnosis and, therefore, no anticancer specific therapy was applied. The median age of the ulcerative colitis patients was 39 years (range, 22–50 years); the gender distribution was 50% female and 50% male. In all ulcerative colitis patients, the disease was active as defined by the colitis activity index >9. Three ulcerative colitis patients received azathioprine (2.5 mg/kg body weight) or 6-mercatopurine (1.5 mg/kg body weight), four patients received 5-aminosalicylic acid, and four patients received corticosteroids. No patient received tumor necrosis factor-α inhibitors. The median age of the healthy control group was 45 years (range, 29–59 years); the gender distribution was 50% male and 50% female. Control patients underwent colonoscopy for colon cancer prophylaxis, and gastrointestinal symptoms such as diarrhea, abdominal pain, or changed stool habits were absent. Signed informed consent was obtained from each subject. Approval of the protocol and consent form was granted by the local ethics committee of the Charité, Berlin, Germany. Biopsies were snap frozen in liquid nitrogen, stored at −80°C, and prepared for MELC analysis as recently described ( 7).

MELC Technology

MELC library. We used a MELC library of 31 fluorescence tags composed of antibodies, lectins, and propidium iodide as a nucleic acid dye as recently published ( 7). The appropriate working dilutions, fluorophore labels, incubation time (15 min), and positions within the MELC run had been established and validated in the course of systematic experiments based on conventional immunohistochemistry and MELC calibration runs ( 8, 9).

Data acquisition by Toponome Imaging Cycler multiepitope readout. MELC technology (EP-patent 0 810 428 and US-patent 6,150,173) has been described previously ( 8, 9). We used a Toponome Imaging Cycler MM3 (TIC) by MelTec GmbH & Co. KG. This fully automated microscopic robot consists of three parts: (a) an inverted wide-field fluorescence microscope (Leica DM IRE2; ×20 air lens; numerical aperature, 0.7) with a cooled charge-coupled device camera (Apogee KX4; Apogee Instruments, Inc.; we use twice binning) and motor-controlled XY stage, (b) a pipetting unit, and (c) a computer that controls all hardware components of the robot using MelTec's TIC-Control Software.

After loading the TIC with the specimen, the robot performed the multiepitope readout as a repetitive cyclic process that consists of three steps: first, the incubation of a fluorochrome-labeled antibody and washing the specimen; second, the acquisition of the fluorescence signal and phase contrast image after correction of any displacement; and third, soft bleaching of the fluorescent dye until no fluorescence signal is detectable. After completion of this incubation-detection-bleaching cycle, the next cycle starts with the next antibody until the whole experimental protocol is completed.

Data analysis and statistics. Fluorescence images produced by each tag were aligned pixel-wise using the corresponding phase contrast images and were corrected for illumination faults using flat-field correction. Finally, pixels not belonging to the biological specimen's information, such as in cases of section artifacts, were excluded as invalid by a mask-setting process. For data reduction purpose, we performed a binarization process. The resulting binarized images composing of the on/off information for the corresponding epitope were superimposed to construct a matrix of combinatorial molecular phenotypes (CMP), which represented a binary code of an epitopes' expression in relation to each pixel (900 × 900 nm² area) of 1,024 × 1,024 pixels. The further analysis dealt with CMP motifs characterizing corresponding pixels. These CMP motifs are defined as pixel-related code of 1/0/* ciphering. The “TopoMiner” software packages (MelTec GmbH & Co. KG) was used for data mining as described by Schubert et al. and Berndt et al. ( 7, 9). Three pairwise screens (Wilcoxon rank-sum test, P < 0.01; search depth, n = 5) were performed to detect differences between the three groups (healthy control, ulcerative colitis, and colorectal cancer).

Data of interest visualization. For visualization of CMP motifs of interest, the “Topolyzer” (MelTec GmbH & Co. KG) was used. This software package allows to visualize CMP motifs of interest as tables, boxplots, toponome maps ( 10), and on the level of fluorescence and binarized images. Colocalization maps were constructed by superimposing a set of illumination-adjusted colorized fluorescence images, with a distinct color and transparency each.

Results

Multidimensional analysis of protein locations in colorectal cancer compared with controls and ulcerative colitis tissue. We previously evaluated and described the reproducibility and robustness of the MELC technology within the intestinal mucosa ( 7). Given the complex changes involved in the generation of colorectal cancer but the limited knowledge about associated changes in the neighboring mucosal layer, we aimed to uncover differences that distinguish colorectal cancer from control tissue and used ulcerative colitis tissue as inflamed control to substantiate specificity of the findings. When mapping 23 surface and 9 intracellular proteins in single tissue samples from 10 control, 10 colorectal cancer, and 10 ulcerative colitis patients, we identified 1,930 CMP motifs being distinct between colorectal cancer and control tissue and 539 CMP being different between colorectal cancer and ulcerative colitis on a significance level of a P value of <0.0005. When calculating the protein expression profiles being 3- and 10-fold different between the different conditions, the high numbers of differentially expressed CMP motifs highlight both shared and unique features in the both entities (Supplementary Table S1). Our study focused on the proteomic changes in the subepithelial lamina propria. Thus, we analyzed cytokeratin-negative and cytokeratin-positive cells separately, and found that 89 distinct motifs were located intraepithelial and 190 motifs in the extra epithelial compartment.

T lymphocytes are increased in the lamina propria of colorectal cancer patients. Given the high number of distinct molecular patterns between colorectal cancer and controls in the lamina propria as well as the importance of the mucosal immune system, we first aimed to gain an overview about the T-cell populations within the intestinal mucosa of control and colorectal cancer patients. The different protein expression patterns shown in control and colorectal cancer tissue by the multidimensional protein localization map substantiate the distinct molecular events observed in both diseases ( Fig. 1A ). Although interesting and novel, these pictures make it obvious that the visual presentation of more than three proteins in one sample results in a picture that is difficult to analyze. In addition, the identification of proteins by conventional immunohistochemistry cannot be quantified, a further limitation of this conventional approach, which can now be overcome by the MELC data mining process. Even more valuable and unique, the MELC technology also allows identifying molecular networks not only by protein colocalization motifs but also the absence of proteins, the so-called anticolocalization code ( 10). Thus, we went on and investigated if all T cells are increased in the colorectal cancer mucosa. As depicted in Fig. 1B, the number of CD3⁺cytokeratin⁻ cells was significantly increased in colorectal cancer comparably to the amount observed in ulcerative colitis tissue. The intestinal mucosa largely consists of CD45R0⁺-memory T cells ( 11); however, in mucosal inflammation such as Crohn's disease, the mucosa is infiltrated by naïve memory T cells from the peripheral blood ( 7, 12). Interestingly, this was not the case in colorectal cancer, where the proportion of CD45RA⁺ and CD45R0⁺ cells in the mucosa was preserved compared with controls (data not shown). Comparably, the CD4⁺cytokeratin⁻ and CD8⁺cytokeratin⁻ cells were equally increased in colorectal cancer, preserving the ratio observed in control tissue (data not shown). However, when screening for cells coexpressing the activation marker nuclear factor-κB (NF-κB), the number of CD8⁺NF-κB⁺ ( Fig. 1C), but not CD4⁺NF-κB⁺ ( Fig. 1D), cells was significantly increased, indicating an increased activation level of CD8⁺ cells in the colorectal cancer mucosa. This tight association with NF-κB seems to be specific for cytotoxic T cells because the number of CD45RA- ( Fig. 1E) and CD45R0-positive cells ( Fig. 1F) coexpressing NF-κB was not increased in colorectal cancer.

CD4⁺CD25⁺ T cells in colorectal cancer. Within the immune system, CD4⁺CD25⁺ T cells play a unique role as regulatory T cells as they inhibit antitumor immune responses ( 13). Indeed, by scanning the subepithelial lamina propria, we could uncover that in colorectal cancer tissue, the number of CD4⁺CD25⁺CD8⁻ cells is decreased ( Fig. 2A ). T cells have to be activated to exert their biological function. We thus wanted to analyze the activation status of T regulatory cells (T_reg) in colorectal cancer and added NF-κB to our panel. As depicted in Fig. 2B and C, in colorectal cancer, CD4⁺CD25⁺CD8⁻ are also not activated, whereas in the inflammatory ulcerative colitis tissue, regulatory T cells are increased.

HLA-expression patterns in colorectal cancer. Distinct patterns of human leukocyte antigens are reported in T cells infiltrating colorectal cancer and adenomas ( 14). Interestingly, on one side, we could confirm this finding by demonstrating that in colorectal cancer, the number of activated CD4⁺, but also CD8⁺NF-κB⁺HLA-DR⁺ T cells, is increased ( Fig. 3A and B ). However, on the other side, this significance was lost when cells were analyzed to be lacking NF-κB coexpression ( Fig. 3C). In contrast, in colorectal cancer, the number of T lymphocytes expressing HLA-DQ and NF-κB were equally contributed in all entities, regardless of their phenotype ( Fig. 3D and E).

Apoptosis of lamina propria T cells and neutrophils. We showed that the number of CD3-positive T cells is increased in colorectal cancer. Given the importance of apoptosis to preserve mucosal homeostasis, we speculated that T-cell apoptosis might be disabled in colorectal cancer and went on to investigate apoptotic pathways. In colorectal cancer, the number of CD3⁺ T cells coexpressing the antiapoptotic bcl-2 and lacking the proapoptotic bax, caspase-3, and caspase-8, is significantly increased compared with controls ( Fig. 4A ). Proteomic analysis revealed that the number of CD4⁺bax⁻bcl2⁺ and CD8⁺bcl2⁺ T cells was increased in colorectal cancer compared with controls ( Fig. 4B and C). Interestingly, when CD45RA and CD45R0 cells were investigated, the number of CD45RA⁺ ( Fig. 4D), but not CD45R0⁺ (data not shown) cells coexpressing the proapoptotic bax, was significantly increased (P < 0.05).

The CD15 antigen is a ligand of endothelial leukocyte adhesion molecules and plays an important role in the adhesion of leukocyte to the vascular wall in inflammation and adhesion between tumor cells and blood endothelial cells ( 15). In our study, the number of CD15-positive cells was equally distributed between normals and colorectal cancer; however, the number of CD15-positive cells was significantly decreased in ulcerative colitis tissue ( Fig. 4E). When analyzing apoptotic pathways, the number of CD15⁺bcl-2⁺caspase8⁻ cells was increased in colorectal cancer compared with controls ( Fig. 4F).

Altered integrin and adhesion molecule expression in colorectal cancer. Integrins regulate adhesion from cells to cells and cells to the extracellular matrix (ECM). It was previously shown that in colorectal cancer, tumor-infiltrating lymphocytes express a lower level of integrin-β1 (CD29) in CD4 and CD8 populations, whereas integrin-α (CD11a) and integrin-β2 (CD18) were reduced in CD8 but not CD4 lymphocytes ( 16). Gating on the cytokeratin⁺CD3⁺ cells, we could confirm the reduction of integrin-β1–bearing cells (data not shown). However, analyzing only cytokeratin⁻ T-cell fraction, the number of CD3⁺CD29⁺ was increased in colorectal cancer compared with controls ( Fig. 5A ). Confirming the seemingly different expression of integrins on colorectal cancer T cells, we could show that only the CD8⁺CD29⁺, but not the CD4⁺CD29⁺ T-cell population, was increased in colorectal cancer ( Fig. 5B and C). This finding remained significant regardless the CD45RA or CD45R0 phenotype (data not shown). The lymphocyte function–associated antigen (LFA)-1 is expressed exclusively on lymphocytes and bind to intracellular adhesion molecule (ICAM)-1 (CD54). In colorectal cancer, the number of CD3⁺ cells with a decreased expression of CD18⁻ as part of the LFA-1 receptor was increased (data not shown). Also, the number of CD3⁺CD11a⁻ was significantly increased in colorectal cancer tissue compared with controls ( Fig. 5D). Cross-checking the results, the number of CD3⁺CD11a⁺ was decreased in colorectal cancer (data not shown), whereas ICAM-1 (CD54) as counterpart of LFA-1 was significantly increased in the subepithelial compartment ( Fig. 5E). The LFA-3/CD2 pathway is crucial to activate mucosal T cells and initiates strong antigen-independent cell adhesion ( 17). We could show that the number of CD3⁺CD2⁻ cells was increased in colorectal cancer, and ulcerative colitis, compared with controls ( Fig. 5F). Fittingly, LFA-3 (CD58) expression is decreased in colorectal cancer tissue ( Fig. 5G). Interestingly, but adding to the significance of the CD2 pathway in mucosal but not peripheral blood T cells ( 18, 19), the inability of T cells to express CD2 in colorectal cancer tissue was only observed in tissue-bound CD45R0 memory T cells ( Fig. 5H) but not CD45RA cells (data not shown).

NK cells. NK cells are a form of cytotoxic lymphocytes that are not only involved in host rejection of tumors but also in inflammation ( 20). By screening for CD56⁺CD3⁻ cells, we could confirm the previous finding that NK cells are increased in colorectal cancer ( Fig. 6A ; ref. 21). Interestingly, we could provide evidence that the CD56⁺CD3⁻ cells coexpressing NF-κB have less apoptotic activity (bax⁻ and caspase-8⁻) in colorectal cancer compared with controls and show decreased coexpression of p53 ( Fig. 6B). NK cells can home in to metastatic sites of gastric and colorectal cancer by up-regulating integrin expression ( 22). When integrin expression profiles were screened on CD56⁺CD3⁻, the number of CD56⁺CD3⁻CD11a⁺ (integrin-α; Fig. 6C), as well as CD56⁺CD3⁻CD29⁺ (integrin-β1; Fig. 6D), but not CD56⁺CD3⁻CD18⁺ (integrin-β2; Fig. 6E), were increased in colorectal cancer, compared with controls and ulcerative colitis. Adding proof to the important role of adhesion molecules, the number of CD56⁺CD3⁻ cells coexpressing CD54 (ICAM-1) is significantly increased in colorectal cancer, compared with both, normal, and ulcerative colitis tissue ( Fig. 6F).

Discussion

System biology approaches seek to integrate high-throughput analyses to elucidate the relationship and interactions between various parts of a biological system. Knowing that proteins, not mRNA, are the functional molecules in the cell, system biology must use proteomic approaches to decipher the molecular mechanism underlying cell signaling ( 23). The novel MELC technology now allows identification of three-dimensional protein associations in selected subcellular compartments and disclose where and when proteins are colocalized in cellular or tissue compartments ( 9). By using this technique, we now mapped functional clusters of proteins and provide, for the first time, context-dependent protein information of the mucosal immune system in colorectal cancer.

Having previously proven the robustness, reproducibility, and representativeness of this novel technique in the human intestinal mucosa ( 7), we tested 7,061,120 motifs of 4.3 × 10⁹ possible CMP motifs and identified nearly 2,000 CMP motifs being distinct between colorectal cancer and control tissue and more than CMP being different between colorectal cancer and ulcerative colitis on a significance level of a P value of <0.0005. The fact that more CMPs are distinctively regulated between colorectal cancer and control tissue than between colorectal cancer and ulcerative colitis might indicate that numerous pathogenic events are shared in both diseases. Thus, by analyzing the proteome topology in colorectal cancer mucosa and comparing it with the healthy but also chronically inflamed mucosa, we aimed to describe and contribute to better understand the host immune response in colorectal cancer and to discriminate between changes due to malignant and inflammatory changes.

It is known that during inflammation, the composition of the cells resident in the intestinal mucosa changes dramatically. We could confirm this finding by demonstrating that in the lamina propria of colorectal cancer patients, the number of T cells is significantly increased up to the level seen in ulcerative colitis tissue. Explaining the increased number of T cells in the lamina propria of colorectal cancer patients, we showed that these cells have an increased expression or lack of antiapoptotic or proapoptotic mediators, respectively. However, the central question arises of whether molecular events during the course of colorectal cancer occur in all cell types comparably, or whether naïve or recently activated cells behave differently than memory or resting cells. Our study revealed that in colorectal cancer, although the proportion of CD4/CD8 and CD45RA/CD45RO cells is equally elevated, their function seems to differ significantly from controls but also ulcerative colitis tissue.

Most known carcinogens and tumor promoters activate the nuclear factor-κB, and thus, it is constitutively active in most tumor cells. On the other side, most agents, including chemotherapeutic agents that induce apoptosis, also induce NF-κB ( 24). In normal resting cells, NF-κB is rarely found to be constitutively expressed but up-regulated in proliferating T cells, B cells, thymocytes, or monocytes ( 25). Thus, in the cytokeratin-negative, subepithelial mucosal compartment, NF-κB expression indicates cell activation ( 24, 25). Having the unique ability to specify and localize protein expression in intact tissue, we showed that in colorectal cancer, not the T-helper population, but the CD8 T-cell population has a significantly increased NF-κB expression. Cytotoxic T cells are mostly CD8⁺, and as essential effectors of the cell-mediated immune response, they can recognize and lyse malignant cells expressing the relevant surface markers ( 26). Their activation in colorectal cancer might uncover an until now unknown autodefense mechanism of the immune system and, thus, fosters the need to closer investigate immunologic changes surrounding the tumor. Pointing to a specificity of this mechanism in colorectal cancer, neither in CD4-, nor in CD45RA-, CD45R0-, or CD56-positive cells, NF-κB expression was increased. As CD8⁺ cells, NK cells have also cytotoxic abilities and are lymphocytes not only involved in inflammation but also host rejection of tumors ( 26). Our study showed that in the colorectal cancer mucosa, NK cells are at an increased level of activation, have less apoptotic activity, and an up-regulated integrin-α, -β2, and ICAM-1 expression, suggesting a distinct role of NK cells in colorectal cancer that needs further evaluation.

T_reg such as CD4⁺CD25⁺ T cells modulate immune responses to self-antigens and foreign antigens ( 27). It was recently shown that in the peripheral blood, mesenteric lymph nodes, and tumor-infiltrating lymphocytes of colorectal cancer patients, T_reg are increased ( 13, 28). Furthermore, CD4⁺CD25⁺ T cells inhibit microbial-induced colon cancer ( 29). T_reg also suppress tumor-specific immunity ( 30) and reduce transendothelial migration of T cells ( 31).

The ability to investigate spatial dissected organs with the MELC technology revealed now that in contrast to the above mentioned immunologic compartments, CD4⁺CD25⁺ T cells are decreased in the mucosa of colorectal cancer patients. When stepping on and investigated why this T_reg population is decreased, we could uncover that these cells are more susceptible toward apoptosis in colorectal cancer. By using NF-κB as an activation marker, we could further show that in the colorectal cancer mucosa, CD4⁺CD25⁺ cells are less activated compared with controls. If the decreased number of T_regs within the mucosa is associated with the increased number of T cells in the colorectal cancer mucosa and permits an increased T-cell response remains speculative. However, the reduced number of inhibitory T cells within the mucosa of colorectal cancer patients might point to a up-regulation of autodefense mechanism of the host mucosa in colorectal cancer.

Further investigating a possible yet underestimated ability of the mucosal immune system to immunologically counterattack the colorectal cancer, we next investigated the role of HLA-DR required for tumor-associated recognition by CD4⁺ cells ( 32). HLA-DR antigen expression on cancer cells correlates with prognosis of colorectal cancer, ( 33) and in our analysis, the number of HLA-DR⁺CD4⁺ was significantly increased in colorectal cancer and ulcerative colitis compared with controls, suggesting that the host increases the antigen recognition in both conditions. Adding evidence to the specificity of this observation, HLA-DR expression was not increased.

Adhesive interactions with the ECM microenvironment are mediated by integrins, a large family of heterodimeric glycoproteins composed of noncovalently associated α and β transmembrane subunits ( 34). T lymphocytes express members of both, the β₁ and β₂, integrin family of cell surface adhesion molecules, which mediate cell-ECM and cell-cell interactions, respectively ( 35). We uncovered that mucosal T cells in colorectal cancer have an increased integrin-β₁ receptor expression. Integrin-β₁ mediates proliferation and inhibits apoptosis of intestinal T cells ( 36, 37), being in line with the increased number of T cells in colorectal cancer mucosa. This link between integrin-β₁ expression and cell expansion can be further corroborate by our observation that in colorectal cancer, the CD8⁺, not the CD4⁺, T-cell population is increased, the same cell type with an increased integrin-β₁ expression profile. In contrast to the increased integrin-β₁ expression, we showed that LFA-1, LFA-3, and ICAM-1 expression is decreased in colorectal cancer. LFA-1 and LFA-3 are expressed on T cells, whereas ICAM-1 is distributed on endothelium as well as on antigen-presenting cells and also in colonic epithelial cells ( 38).

In conclusion, by using a novel, automated, multidimensional, fluorescence-based microscopy robot technology, we performed the first proteomic analysis of the intestinal mucosa in colorectal cancer patients. Our study showed that with regard to the local immune system, the presence of colorectal cancer induces a tremendous modification of protein expression profiles within the lamina propria. The analysis of key immune function–related proteins and in situ detection of their modification may not only facilitate our understanding of the pathogenic events observed in colorectal cancer but also might help to understand how the host responds to treatment. As the adaptive immune system rapidly emerges as a major player in tumorigenesis ( 39), the assembly of a toponomic picture in colorectal cancer by the novel MELC technology is a “work in progress” that might add great value to optimize preclinical therapy studies and, thus, aid the development of novel antitumor immunotherapy approaches in colorectal cancer.