Article Text

Download PDFPDF

Prognostic impact of β-2-microglobulin expression in colorectal cancers stratified by mismatch repair status
  1. Viktor Hendrik Koelzer1,
  2. Kristi Baker2,
  3. Daniela Kassahn1,
  4. Daniel Baumhoer3,
  5. Inti Zlobec1
  1. 1Institute of Pathology, University of Bern, Bern, Switzerland
  2. 2Department of Gastroenterology, Brigham and Women's Hospital, Boston, Massachusetts, USA
  3. 3Institute of Pathology, University Hospital Basel, Basel, Switzerland
  1. Correspondence to: Inti Zlobec, Translational Research, Institute of Pathology, University of Bern, Murtenstrasse 31, Room 414, Bern CH-3010, Switzerland; inti.zlobec{at}


Background β-2-microglobulin (B2M) is essential for antigen presentation, yet may also possess proto-oncogenic properties.

Aim To determine the prognostic impact of B2M in patients with mismatch repair (MMR) proficient and deficient colorectal cancer (CRC) and to investigate whether this effect on outcome is dependent on the local immune response.

Methods B2M protein expression and tumour-infiltrating immune cells (CD3, CD16, CD163, CD20, CD4, CD45RO, CD56, CD68, CD8, FoxP3, GranzymeB, iNOS, mast cell tryptase, MUM1, PD1, TIA-1) were evaluated in a well characterised tissue microarray of 408 CRCs. The predictive value for clinicopathological features and the prognostic significance of B2M expression were analysed, stratified by MMR status and the immunohistological characteristics of immune cell infiltrates.

Results Interobserver agreement for B2M staining was high (intra-class correlation coefficient=0.91). Complete B2M loss was more frequent in MMR-deficient (19.4%) compared to MMR-proficient (7.1%) tumours (p<0.001). In MMR-deficient cases, B2M loss predicted rare local recurrence (p=0.034), infrequent nodal-positivity (p=0.035), absence of distant metastasis (p=0.048; sensitivity=100%) and a trend towards favourable survival (p=0.124) independent of immune infiltrates. No associations between B2M and clinicopathological features were observed in MMR-proficient cases.

Conclusions Our data show for the first time that absence of B2M protein expression identifies MMR-deficient cancers with a favourable clinical course and absence of metastatic disease. Validation of B2M protein expression for sub-classification of MMR-deficient CRC is recommended for future clinical trials.

Statistics from


Colorectal cancers (CRCs) can be classified by the ability of the DNA mismatch repair (MMR) machinery to restore defects within microsatellite repeats.1 MMR-deficient tumours comprising 15% of all carcinomas can occur in both the sporadic (10–12%) and hereditary (3–5%) setting and are characterised by methylation or mutation of MMR genes leading to gene silencing and high-level microsatellite instability (MSI-H). Immunohistochemical analysis for MMR proteins (MLH1, MSH2, MSH6 and PMS2) is used in diagnostic pathology to identify MSI-H CRCs, though no distinction between sporadic and hereditary cases can be made by protein expression alone. The remaining 85% of cases often arise from mutations in APC, KRAS and p53, show chromosomal instability, retain expression of MMR proteins and are thus MMR-proficient.2

MMR-proficient and -deficient cases differ in terms of their clinical, histomorphological and molecular features.3 ,4 In general, MMR-deficient patients have a more favourable survival time and infrequent distant metastatic disease in comparison to their MMR-proficient counterparts.5 It has been acknowledged that immune surveillance may efficiently control the outgrowth and systemic spread of MMR-deficient tumours, especially in patients with germline mutations leading to the hereditary non-polyposis CRC (HNPCC, Lynch syndrome).6 ,7 Indeed, recent studies show that the frameshift mutations resulting from MMR-deficiency can lead to the generation of tumour-associated frameshift-peptides, thus eliciting specific T-cell responses.7 However, the stringent immune selection process and immunoselective pressure on MMR-deficient tumours may also contribute to an outgrowth of less immunogenic cancer cell phenotypes with an increased frequency of HLA class I defects.8–10 Truncating mutations in coding microsatellites of the β-2-microglobulin (B2M) gene have been identified as the primary mechanism impairing HLA-class I antigen presentation, leading to the generation of immune escape variants.11 It should be expected that these immune escape variants of MMR-deficient tumours generally exhibit a more aggressive cancer phenotype with more frequent metastasis and unfavourable clinical course if the primary consequence of B2M loss is interruption of immune control.

Unexpectedly, retrospective correlative studies and a recent well designed prospective clinical trial have provided conclusive evidence that B2M mutation status may serve to identify colon cancer patients with favourable clinical course.9 ,12 In particular, B2M-mutant MMR-deficient cancers have been characterised by lack of distant metastasis and disease relapse.9 ,12 Interestingly, experimental evidence from different cancer models suggests that B2M may act as a proto-oncogenic factor promoting tumour cell growth, epithelial mesenchymal transition (EMT), migration and systemic spread.13–15 Furthermore, serum levels of free B2M show a significant correlation with tumour load in several carcinomas as well as haematopoetic malignancies, and free B2M has been suggested as an independent biomarker for tumour progression.16–18 This suggests that B2M loss in MMR-deficient CRC may not only impair immune recognition of MHC-bound antigen on the tumour cell but at the same time may entail loss of an important oncogenic factor for systemic spread.

We aim to specifically address this dichotomy between the immune and proto-oncogenic function of B2M in CRC by systematically analysing the prognostic and predictive value of B2M protein expression in correlation with immune cell infiltrates in a well characterised patient collective. Further, we hypothesise that B2M protein expression may be a valuable prognostic indicator for sub-classification of MMR-deficient CRC by immunohistochemistry.

Materials and methods


A total of 408 non-consecutive patients with primary CRCs (including 98 rectal cancers; 24.7%) diagnosed at the Triemli Hospital, Zurich, Switzerland between 1988 and 1996 were entered into this study. Histopathological criteria including TNM stage (6th edition, UICC) was reviewed by an experienced gastrointestinal pathologist. From clinical reports, information on survival, distant metastasis, local recurrence as well as postoperative therapy was available. Clinicopathological data are shown in table 1. Mean survival time was 64.3 months. No patients received preoperative therapy.

Table 1

Characteristics of patients in the analysis cohort with evaluable β-2-microglobulin (n=408) stratified by mismatch repair (MMR) status


Tissue microarray

A tissue microarray (TMA) of formalin-fixed paraffin-embedded tissue blocks from these 408 patients was constructed.19 Seventy-one tissue samples from normal colorectal mucosa were included as a control. Immunohistochemistry for MMR proteins MLH1 (clone MLH-1; dilution 1:100; BD Biosciences Pharmingen, San Jose, California, USA), MSH2 (clone MSH-2; dilution 1:200; BD Biosciences Pharmingen) and MSH6 (clone 44; dilution 1:500; BD Biosciences Pharmingen) was performed19 and tumours were classified into two groups:20 (1) MMR-proficient cases expressing MLH1, MSH2 and MSH6; and (2) MMR-deficient cancers demonstrating an absence of expression of one of more of these proteins. Since the aim of this study was to evaluate differences in B2M expression by MMR status, no distinction between likely sporadic or hereditary MMR-deficient cancers was undertaken.

Immunostaining for B2M and immune markers was performed according to standard protocols (table 2) and was evaluated by two experienced pathologists. B2M expression was predominantly found on the tumour cell membrane and also within the cytoplasm. Cases were classified into complete loss (0%) or any positive staining (>0%). For tumour infiltrating immune cells, the total number of reactive cells within the tumour microenvironment was evaluated independent of localisation (intratumoural or stromal). For PD1 and iNOS, cases were scored as the absence or presence of any positive cells.

Table 2

Immunohistochemistry protocols

Whole tissue sections

In order to verify the homogeneous negative or positive B2M staining throughout the tumour, 10 primary CRCs were randomly selected from the archives of the Institute of Pathology, University of Bern, Switzerland. Paraffin-embedded tissue blocks were cut at 4 μm. Immunohistochemistry for B2M was performed (Dako, 1:100) and negative controls were included for each case. Ten out of 10 tumours were homogeneously positive or negative for B2M, including five negative and five positive cases.

Statistical analysis

Inter-observer agreement was assessed using the intra-class correlation coefficient (ICC) and κ statistic. Values greater than 0.8 indicate excellent agreement between independent observers’ scores. The optimal cut-off score for classifying cases as B2M positive was determined using receiver operating characteristic curve analysis. Categorical variables were analysed with the χ2 or Fisher's exact test. Kaplan–Meier curves and log-rank tests were used to determine survival time differences. Spearman's correlation was used to determine the correlation between B2M staining and number of positive tumour infiltrating lymphocytes. The independent prognostic effect of B2M was analysed using a multiple Cox regression model and effect size was determined using HRs. A power calculation using the following parameters was performed to test the relevance of the associations between B2M and distant metastasis or local recurrence (p1: proportion of patients with loss of B2M and the outcome; p2: proportion of patients without loss of B2M and the outcome; relative sample size; and significance level α=0.1; one-sided). The power for both analyses was determined to be 77% for distant metastasis and 74% for local recurrence. Analyses were carried out using SAS V.9.1.


Inter-observer agreement of B2M staining

The inter-observer reproducibility of the B2M staining was assessed across the whole range of values (0–100%). The ICC, which takes into consideration the between-subject and within-subject variances, was 0.91, indicating excellent agreement between observers’ scores. Percent concordance between B2M loss (0% tumour cells stained) and B2M positivity (>0% tumour cells stained) was 96.2% which corresponded to a κ=0.84 (95% CI 0.76 to 0.92), again indicating excellent inter-observer agreement. Representative photo-micrographs of B2M staining on both the TMA and whole tissue sections are shown in figure 1.

Figure 1

Representative photomicrographs of tissue microarray punch (50×) with (A) positive and homogeneous β-2-microglobulin (B2M) staining in the membrane and cytoplasm of colorectal cancer (CRC) cells and (B) homogeneous loss of B2M staining. Whole tissue section of CRC (20×) showing (C) homogeneous and positive B2M staining with cytoplasmic and membranous expression and (D) homogeneous loss of B2M expression with positively stained intra-tumoural lymphocytes (arrow). This figure is only reproduced in colour in the online version.

Stratification of patients by MMR status

Of the 482 patients with full clinicopathological data, B2M could be evaluated in 408 CRCs. Patient characteristics are shown in table 1 and underline the expected differences between MMR-deficient and -proficient tumours with respect to female gender (p<0.001), right-sided tumour location (p<0.001), more advanced pT classification (p<0.001), higher tumour grade (p=0.038) and absence of local recurrence (p=0.005).

B2M staining in normal mucosa, MMR-deficient and -proficient CRCs

In total 41/408 CRCs (10%) and 1/34 (2.9%) normal colonic mucosa tissue showed a loss of B2M protein expression (p=0.085). We observed 19/98 (19.4%) cases with complete B2M loss in the MMR-deficient compared to 22/310 (7.1%) in the MMR-proficient carcinomas (p<0.001) (figure 2A).

Figure 2

(A) Difference in the number of tumours with loss or positive staining for β-2-microglobulin (B2M) stratified by mismatch repair (MMR) status. (B) Differences in the presence or absence of distant metastasis with B2M expression stratified by MMR status. (C) Differences in pN classification stratified by B2M expression and MMR status. (D) Kaplan–Meier curve showing the more unfavourable survival in patients expressing B2M in the MMR-deficient setting; y-axis represents the percentage of patients still at risk of death. This figure is only reproduced in colour in the online version.

Prognostic effect of B2M loss in MMR-deficient and -proficient cancers

A shown in table 3, in MMR-deficient tumours (n=98), complete loss of B2M was associated with mucinous histological type (p<0.001), higher tumour grade (p=0.019) and absence of local recurrence (p=0.034). Interestingly, none of the 19 (100%) MMR-deficient cases with complete loss of B2M presented with, or eventually developed distant metastasis (p=0.048; figure 2B). A significant association between B2M positivity and higher pN stage was observed (p=0.035). In particular, the frequency of lymph node positivity was 1.85 times higher in patients with B2M positive tumours compared to those with loss of the protein (figure 2C). A trend towards prolonged survival with an average 5-year survival rate of 91.7% was observed among cases with MMR deficiency and complete loss of B2M protein expression in comparison to 72.1% in patients with positive staining for the protein independent of postoperative therapy; however this did not reach statistical significance (p=0.124; figure 2D). In MMR-proficient cases (n=310), no associations between staining of B2M, clinicopathological features or survival time were noted.

Table 3

Association between complete loss of β-2-microglobulin (B2M) staining and clinicopathological features in mismatch repair (MMR) proficient and deficient colorectal cancers

Correlation of tumour expression of B2M and tumour infiltrating lymphocytes

Next, we correlated staining of B2M with the presence of different TIL populations. B2M positivity showed a moderate positive correlation (r>0.2) with the presence of CD3 (r=0.24), CD16 (r=0.26), CD8 (r=0.34), GranzymeB+ (r=0.21) and TIA-1+ (r=0.29) tumour infiltrating cells in MMR-deficient tumours, and with CD3 (r=0.27), CD8 (r=0.28), GranzymeB+ (r=0.29), MUM1+ (r=0.23) and TIA-1+ (r=0.22) positive cells in MMR-proficient cases. No association of B2M expression with CD4, CD20, CD45RO, CD56, CD68, CD163, FoxP3+, iNOS+, mast cell tryptase+, MUM1+ and PD1+ immune cells was observed.

In MMR-proficient cases B2M expression showed a moderate positive correlation with CD3 (r=0.27), CD8 (r=0.28), GranzymeB+ (r=0.29), MUM1+ (r=0.23) and TIA-1+ (r=0.22) positive cells. For CD4, CD16, CD20, CD45RO, CD56, CD68, CD163, FoxP3+, iNOS+, mast cell tryptase+ and PD1+ immune cells, no association with B2M expression was observed.

Survival time analysis of lymphocyte populations in MMR-deficient cancers

None of the evaluated lymphocyte markers showed a correlation with survival time in MMR-deficient cancers: CD3 (p=0.785), CD16 (p=0.449), CD163 (p=0.551), CD20 (p=0.542), CD4 (p=0.906), CD45RO (p=0.182), CD56 (p=0.509), CD68 (p=0.286), CD8 (p=0.502, exemplary stain shown in figure 3A), FoxP3 (p=0.309), GranzymeB (p=0.247), iNOS (p=0.943), mast cells (p=0.86), MUM1 (p=0.094), PD1 (p=0.786) and TIA-1 (p=0.723). Of all these markers only the presence of CD8 cells was more frequent in patients with no metastatic disease (p=0.0106). Next, we determined whether the predictive value of B2M expression for the absence of metastasis was independent of CD8 cells. Indeed in both CD8 and CD8− subgroups, sensitivity for absence of metastasis was 100% in patients with loss of B2M staining. Multivariate survival time analysis for B2M and CD8 showed a significant effect of B2M positive expression with poorer outcome (HR 2.34, 95% CI 1.04 to 5.24, p=0.0396), suggesting its effect is independent of CD8 expression (figure 3B–D).

Figure 3

(A) Expression (left) and lack of expression (right) of CD8. Kaplan–Meier curves showing (B) no difference in survival time in mismatch repair (MMR) deficient patients stratified by CD8 expression. More favourable survival time with β-2-microglobulin loss in patients with (C) CD8 MMR-deficient and (D) CD8− MMR-deficient cancers; y-axis represents the percentage of patients still at risk of death. This figure is only reproduced in colour in the online version.


The novel findings of this study show that complete loss of B2M protein expression may have significant prognostic implications for patients with MMR-deficient CRCs. Loss of the protein, detected by immunohistochemistry, corresponds to significantly fewer lymph node positive cases, few local recurrences and lack of systemic spread.

In a first step, we found that MMR-deficient CRCs had significantly more frequent B2M protein expression loss (19.4%) compared to MMR-proficient tumours (7.1%). This finding is in line with data from Kloor and Yamamato, who showed that MMR-deficient CRCs frequently display mutations of the B2M gene in 15.4–50% of cases with an elevated frequency in HNPCC-associated MSI-H CRC.9–11 Bernal et al21 further characterise the phenotype of MMR-deficient CRC with B2M mutations with absence of HLA class I protein and fewer lymphocytic infiltrates, and postulate that the increased frequency of HLA-I negative tumours may be a consequence of immunoselection.

Second, we report that none of the MMR-deficient cancers with loss of B2M protein expression experienced distant metastases and the rate of local recurrence was decreased among these patients. Kloor et al9 ,22 have suggested that mutation of B2M may promote local tumour progression but also protect against formation of distant organ metastases in patients with MMR-deficient CRCs, particularly in the hereditary HNPCC syndrome. The authors speculate that maintained expression of B2M or of assembled HLA class I complexes may favour the formation of distant metastasis. Indeed, recent studies document that B2M may possess pleiotropic functions similar to a typical proto-oncogene and promote EMT, cell migration and invasion, and metastatic spread of epithelial malignancies.13–15 ,18 A role of B2M expression as a proto-oncogenic factor in CRC is supported by positive staining of liver metastases with HLA class I antigen in more than 95% of cases.23 In line with these findings, Kloor et al document a complete absence of metastasis in patients with mutations of B2M in a collective of 104 MMR-deficient CRCs, and provide further evidence of B2M as a prospective marker for absence of disease relapse in the 223 patients of the FOGT-4 collective.9 ,12 Our results are in agreement with these previous data showing no cases with distant metastasis in patients with MMR-deficient tumours and B2M mutations.

In a third step, we show that the loss of B2M expression identifies a subgroup of patients with MMR-deficient CRC with favourable prognosis. We show that this prognostic effect of complete loss of B2M expression is confined to patients with MMR-deficient tumours. Although prognostic studies on B2M protein expression are few, several works document the effect of MHC class I molecules on outcome. Menon et al24 found that down-regulation of HLA-A corresponds to lower tumour stage, mucinous cancers, lower recurrence rates and longer disease-free survival time in 88 CRC patients. All patients with complete loss of HLA-A had no recurrence and half were MSI-H. Additionally, Blum et al25 found that patients with tumours expressing high ratios of microtubule associate protein 7 to B2M had significantly more favourable outcome. In the FOGT-4 clinical trial Tikidzhieva et al12 have also provided first prospective evidence that B2M mutations are associated with a trend towards absence of tumour related death events in MMR-deficient patients. In line with these findings, we show here that complete loss of B2M protein expression correlates with lower rates of local recurrence in addition to prolonged disease-specific survival time in MMR-deficient patients only.

In this study, we show only a negligible correlation between B2M staining and tumour-infiltrating CD8 cytotoxic T-lymphocytes and CD56 natural killer cells in MMR-deficient cancers.26 ,27 Defects in the antigen presentation and processing machinery or down-regulation of HLA class I molecules have been reported to be more frequent in MSI-H compared to microsatellite-stable CRCs including mutations in B2M, as well as in TAP1 and TAP2, two genes involved in the transport of peptides to the endoplasmic reticulum.28 ,29 Kasajima and co-workers found that the presence of TAP1 and TAP2 by immunohistochemistry was significantly correlated to the presence of CD4 and CD8 cells but that only these tumour-infiltrating lymphocytes and not TAP1, TAP2 or MHC class I had an effect on prognosis.30 In this study, we find that expression of B2M correlates moderately with the expected increases in intratumoural lymphocytes involved in cytotoxic T-cell mediated responses, namely CD3, CD8, GranzymeB+ and TIA-1+ irrespective of MMR status. However the presence of these cell types had no effect on prognosis in MMR-deficient patients. This may be due to decreased immune recognition of MHC-deficient tumour cells.28 Indeed a prognostic effect of CD8 T-cell infiltration in MMR-deficient CRC has not been observed.31

Especially in stage II CRC, it remains a challenge to identify patients that may benefit from adjuvant chemotherapy. Our findings suggest that B2M possesses diagnostic value to classify MMR-deficient cases into high- and low-risk groups and should be prospectively validated. In a first step, we report that B2M is a highly reproducible biomarker between different observers in accordance to the REMARK guidelines,32 and is homogeneously expressed and lost throughout CRCs. This suggests that biopsy material may also be suitable for the analysis of B2M expression, and that the assessment of positive staining (>0%) should lead to few challenges in interpretation in the diagnostic routine.

This is the first study to document a predictive value of B2M protein expression in MMR-deficient CRCs by immunohistochemistry for the absence of metastatic spread in a large cohort of patients with full TNM stage, follow-up and therapy information. This supports the hypothesis that B2M mutations may have a protective function against the development of metastatic disease that is unrelated to the immune response. Although prospective studies need to be performed to validate these findings, our preliminary data suggest that B2M protein expression may be useful as a diagnostic marker to identify high- and low-risk patients with MMR-deficient cancers who may benefit from closer follow-up.

Take-home messages

  • This is the first study to document a predictive value of β-2-microglobulin (B2M) protein expression in mismatch repair-deficient tumours by immunohistochemistry for the absence of metastatic spread.

  • Our results support not only the hypothesis that B2M function on tumour cells may not be restricted to antigen presentation as part of the MHC-I complex, but also that B2M may be a useful marker for identifying high-risk MMR-deficient CRC patients.


The authors would like to thank Professor Alessandro Lugli for critical review of the study and Professor Alexandar Tzankov for his contribution regarding analysis of immunohistochemical markers.



  • Contributors The six authors are justifiably credited with authorship, according to the authorship criteria. In detail: VHK: acquisition, analysis and interpretation of data, drafting of the manuscript, final approval given; KB: interpretation of data, critical revision of manuscript, final approval given; DK: critical revision of manuscript, final approval given; DB: critical revision of manuscript, final approval given; IZ: conception, design, analysis and interpretation of data, drafting of the manuscript, final approval given.

  • Funding Molecular Pathology Division of the Institute for Pathology, University Hospital Basel.

  • Competing interests None.

  • Ethics approval Ethics Committee of the University Hospital of Basel (Ethikkomission Basel).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.