Co-expression of HLA-I loci improved prognostication in HER2+ breast cancers

Introduction

The studies on tumor immunology have revolutionized our understanding of cancer pathogenesis and treatment. Immunotherapy for cancer management is now widely adopted in various cancer types. Durable responses and improved survival had been observed in patients treated with immune checkpoint blockade across a wide range of cancers [1]. Currently immunotherapy has been approved for melanoma, non-small-cell lung cancer, head and neck cancer, renal cell carcinoma and bladder cancer [1], and very recently in advanced triple negative breast cancer [2]. Despite that, biomarker development for accurate prediction of treatment response is less successful as cancer immune interaction is multifactorial. A better understanding of anti-tumor immunity is important to decipher the underlying mechanism of cancer immunosuppression and aid in biomarker development.

Tumor-associated antigens presenting by major histocompatibility complex (MHC) class I molecules to cytotoxic T cells is fundamental to anti-tumor immunity. MHC class I molecules are encoded by genes located within the MHC region on chromosome 6 and comprise three classical human leukocyte antigens (HLA-I), namely HLA-A, -B and-C in humans. In normal cells, six different HLA alleles, two from each locus, are expressed. These HLA alleles form a tri-molecular complex with β2-microglobulin and peptide antigen on the cell surface for antigen presentation. Loss of HLA-I expression is a frequent event in different cancers for immune evasion, allowing tumor dissemination and metastasis [3]. Several mechanisms of HLA loss have been proposed, leading to different patterns of HLA-I expression on cell surface. Depending on whether the genomic deletion and epigenetic regulation affect the entire HLA-I region, single locus or specific allele, HLA-I can be totally or partially absent from tumor cells. They can be downregulated by alterations in its regulatory signaling pathways and defects in the assembly and stability of HLA complex, such as downregulation of antigen processing machinery and β2-microglobulin. These can lead to loss of expression in all HLA-I loci [4].

In breast cancers, downregulation of HLA-I expression is common, being observed in 47–85% of primary tumors [5–13]. High HLA expression was consistently associated with higher tumor grade [5, 8, 9, 11], and variably with hormonal receptor negativity [5], HER2 negativity [13], lymphovascular invasion [9], and tumor infiltrating lymphocytes (TIL) [9]. Its relationship with outcome was also variable [5, 6, 8, 11, 13]. Most studies used antibodies that recognized the non-polymorphic region of HLA-I, thus did not differentiate the expression of the different HLA loci [5–13]. It is possible that alterations in specific locus could be observed with different clinical ramifications. In fact, in colon carcinoma, only alterations in HLA-A and C had prognostic relevance [14]. In endometrial cancer, loss of HLA-B and C, but not HLA-A, was associated with deficient mismatch repair [15].

In this study, we examined HLA-I expression in breast cancer using antibodies reactive to different HLA-I loci. The expression of the different HLAs and their co-expression status were assessed with various clinico-pathologic features, biomarker expression and patients’ survival. In addition, analysis was performed on paired primary tumors and corresponding nodal metastases to examine the relationship of HLA expression with tumor dissemination.

Materials and methods

Results

In total, 1190 primary invasive breast cancers and 314 paired metastatic lymph nodes were included in this cohort. The mean patients’ age at diagnosis was 54.4 ± 12.7 years (range 22–97 years) and the mean tumor size was 2.69 ± 1.55 cm (range 0.2–13.0 cm). Of the 1190 cases, 175 (14.7%), 477 (40.1%), and 538 (45.2%) were of grade I, II, and III respectively. ER, PR, and HER2 were positive in 70.3% (832/1183), 66.2% (780/1179), 19.0% (225/1184) of the cases respectively. High HLA-A, B and C expression was found in 27.8% (311/1120), 33.3% (380/1141) and 40.1% (452/1126) respectively. Among the 1054 cases with data on all three HLAs, 698 (66.2%) showed concordant expression (including 509 (48.3%) cases with low expression for all three markers (all low) and 189 (17.9%) cases showed all HLA high (all high) expression). The remainder (33.8%; 356/1054) expressed mixed levels of the three HLAs (mixed low/high). (Details on the combined expression of the HLA-A, B and C were shown at supplementary table S2).

Discussion

Evasion of immune destruction is one of the important traits that promotes tumor transformation. The recently developed immunotherapies aim to overcome the immune escape and reinvigorate the anti-tumor immune function. However, tumor regression can only be achieved in a proportion of patients, indicating that treatment outcome may be determined by both the immunologic features of the host as well as the immune evasion mechanism of the tumor cells.

One of the fundamental mechanisms by which many cancers, including breast cancer, evade immune surveillance is by down-regulation of their HLA-I expression [4]. Low expression in all three HLA-I loci was observed in the current cohort, with the expression rates of HLA-A, B and C being 28.9%, 33.7% and 39.8% respectively. Expression of individual HLA-I loci was associated with features of high grade breast cancers, including the presence of necrosis, high TIL level, negative hormonal receptor expression, higher Ki67, basal markers and pAKT. Similar association of HLA-I expression with high grade tumor were also reported previously [8, 9]. In line with that, IFNγ signaling is reported to be more robust in breast cancer lacking PR [17]. Although some earlier reports with small cohort or using specific cell model suggested an inverse correlation of HLA-I with HER2 expression [18, 19], in our study as well as some others [5, 7–10], this was not consistently observed. In addition, there appeared to be a lack of association of HLA-B expression with growth receptor expression. The difference could be due to the additional SP1 regulatory elements in HLA-B locus. HER2 signaling could lead to phosphorylation of SP1 and later its transcriptional activity [20].

All HLA low expression was found in 48.3% of the cases. Only 17.9% showed all high HLA expression and 33.8% showed mixed low/high expression of down-regulation of at least one HLA-I locus. Previous studies mainly used anti-pan HLA antibodies [5–13]. Our results from co-expression analysis agrees with the previously reported down-regulation rate of 47% or more. However, these earlier studies did not differentiate between tumors with mixed low/high and all high HLA expression. We observed different clinico-pathologic features and, importantly, prognostic implication between cases with different co-expression patterns. In fact, each HLA-I allele was associated with different peptide binding repertoires [21]. Loss/down-regulation of any HLA-I locus would be predicted to present a smaller and less diverse repertoire of tumor-derived antigen to immune cells as compared to those with high expression in all loci; thus with a smaller activated T cell repertoire. In addition, selective loss/ downregulation of HLA loci could result in both downregulation of T cell and NK cell mediated tumor cytotoxicity. HLA-I could be ligands for inhibitory NK cell receptors that signal NK cells to remain silent to self antigen. Total loss of HLA can potentially allow tumor cells to become susceptible to NK cell killing. However, in cases with selective HLA loci downregulation (i.e. mixed low/high phenotype), the tumor cells may escape from both T cell and NK cell mediated cytotoxicity. In line with that, recent data demonstrated a progressive decrease in pCR rate from HLA-negative to HLA-low and HLA-normal HER2+ cases while HLA-high group showed the highest response [22]. Compared to mixed high/low cases, cases with all high expression demonstrated the best survival in the HER2+ subset and were associated with a higher level of TIL. All high expression may represent an intact antigen presenting mechanism; thus enabling an efficient activation of anti-tumor immunity and control of tumor progression. Significance of prognostic value for assessing the co-expression pattern of HLA loci could be suggested.

It is interesting to note that the prognostic value of the HLA co-expression could be affected by the TIL level and HLA expression pattern on nodal metastases. In HER2+ cancers, high TIL predicted benefit of anthracycline chemotherapy, trastuzumab treatment and better prognosis [23–25]. Of note, only those cases with high TIL together with homogeneous HLA high expression in our cohort demonstrated good prognosis while those with low TIL or low/ heterogeneous HLA had poor survival. The high TIL level indicated the presence of immune surveillance. However, to elicit the effector functions, antigen presentation by HLA expressed on tumor cells is an important pre-requisite. Therefore, the prognostication of TIL should be evaluated in the context of HLA status of the tumor cells. Combined assessment of TIL and HLA expression may provide more accurate prognostic and predictive information. Moreover, downregulation of HLA expression in nodal metastases compared to primary tumor was observed in this study. Favorable outcome was found with cases showing high expression of HLA at both sites, in particular, cases with all high HLA expression in all three loci. Nodal metastasis is one of the most important indicators of poor prognosis, as it is thought that cancer cells from metastatic lymph nodes can further escape into the circulation by invading nodal blood vessels leading to the development of distant metastases [26]. Downregulation of HLA or antigen presenting machinery on metastatic tumors has long been described in metastatic breast cancers [27–29]. The current results fit the concept that HLA loss may favor metastatic colonization. Thus, a better outcome could be observed in tumors maintaining high HLA expression in both primary tumor and nodal metastases. This HLA expression pattern could potentially identify node positive patients with better outcome.

Interaction between tumor cells and the immune microenvironment is of pivotal importance in inducing an effective anti-tumor immune response. The loss of HLA-I expression on cancer cells could be a hurdle for successful immunotherapy. It has become increasingly evident that HLA-I phenotype could determine the clinical success of immune checkpoint blockade. Mutations in IFNγ signaling pathway and β2-microglobulin which are required for HLA expression or upregulation were associated with acquired resistance to PD-1 blockade in melanoma [30]. In a cohort of melanoma patients treated with immune checkpoint blockade (ICB), patients with loss of heterozygosity (LOH) or homozygosity in at least one HLA-I locus had reduced survival, independent of somatic mutational load, tumor stage, age or types of treatment [31]. Success of antitumor response unleashed by ICB could depend on the efficacy of tumor antigen presentation to immune cells. An intact HLA phenotype is required for the presentation of a diverse repertoire. Tumor mutational burden (TMB) has been associated with ICB benefit and is currently considered to be a potential biomarker for ICB [32]. However, a higher TMB in tumors with HLA LOH was observed compared with those without HLA LOH in lung cancer patients enrolled in the TRACERx study [33]. One could speculate that patients with a high TMB and HLA LOH may not respond equally to ICB as those without LOH. All these indicated that HLA-I status/expression should be taken into account to predict treatment response. In term of therapy, strategies to restore HLA-I expression for efficient antigen presentation can be helpful. The immunogenic modulation on tumor immune environment, such as up-regulation of inflammatory cytokines and Fas expression as well as an increased MHC I molecules, by low-dose chemotherapy and radiation have been suggested [34, 35]. In a recent clinical trial for metastatic TNBC, induction treatment with short-term chemotherapy or irradiation followed by ICB led to clinical benefit in a substantial subset of patients, with higher than expected response rate and durable responses [36]. Further studies exploring the relationship of HLA and ICB are warrant to provide scope for improved treatment strategies and prediction of treatment responses.

Limitations of the study included the choice of locus specific antibodies. Although the antibodies were raised against specific locus and their application has been verified in other studies, it is possible that they may recognize some specific alleles in the same locus with the lower efficiency due to variability of the regions. Further characterization of the antibodies with a full panel of HLA-I alleles will be required to examine the antibody sensitivity. Nonetheless, the study provide some initial data deserve further validation for its potential clinical utility.

In summary, a combined analysis on the co-expression status of HLA-I loci may be important for the prognostication of HLA expression. It can be incorporated into TIL evaluation and nodal status to provide improved prognostication. In addition, a better outcome could be observed in tumors maintaining high HLA expression in both primary tumor and nodal metastases highlighting the importance of its evaluation on both primary and metastatic tumors. In the era of immunotherapy, HLA-I could to be an important component to be considered. Strategies to promote and assess its expression may be essential.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

JT analyzed the data and wrote the paper; CH, YS, IP, MM performed the experiments; S-KC, S-YC, KS collected and arranged clinicopathological data of cases; GT conceived the idea for the paper, provided guidance and critically revised the paper. All authors read and approved the final version to be published.

Compliance of ethical standards