HLA class I loss in colorectal cancer: implications for immune escape and immunotherapy

Abstract

T cell-mediated immune therapies have emerged as a promising treatment modality in different malignancies including colorectal cancer (CRC). However, only a fraction of patients currently respond to treatment. Understanding the lack of responses and finding biomarkers with predictive value is of great importance. There is evidence that CRC is a heterogeneous disease and several classification systems have been proposed that are based on genomic instability, immune cell infiltration, stromal content and molecular subtypes of gene expression. Human leukocyte antigen class I (HLA-I) plays a pivotal role in presenting processed antigens to T lymphocytes, including tumour antigens. These molecules are frequently lost in different types of cancers, including CRC, resulting in tumour immune escape from cytotoxic T lymphocytes during the natural history of cancer development. The aim of this review is to (i) summarize the prevalence and molecular mechanisms behind HLA-I loss in CRC, (ii) discuss HLA-I expression/loss in the context of the newly identified CRC molecular subtypes, (iii) analyze the HLA-I phenotypes of CRC metastases disseminated via blood or the lymphatic system, (iv) discuss strategies to recover/circumvent HLA-I expression/loss and finally (v) review the role of HLA class II (HLA-II) in CRC prognosis.

Introduction

It is well established that the loss of human leukocyte antigen class I (HLA-I) in tumours is a major escape mechanism from T cell-mediated responses.^1–5 This biological phenomenon observed during the natural history of tumour development has also important implications when T cell-mediated immunotherapy is applied in cancer patients using a variety of procedures,^6,7 including Bacillus Calmette–Guérin (BCG) in bladder cancer,⁸ immunization with ‘unique’ or ‘shared’ tumour antigens,^9,10 adoptive T cell transfer in melanoma¹¹ or the use of monoclonal antibodies (mAbs) that target molecules involved in the regulation of T cell responses such as anti-PD-1/PD-L1.¹² It is therefore mandatory to analyze in detail the expression of HLA-I in primary tumours and in metastases.

HLA-I and -II molecules play a pivotal role in presenting processed viral or tumour antigens to CD8+ and CD4+ T lymphocytes as short peptides. Primary tumour lesions are in origin HLA-I positive. T cell infiltration of the tumour nest is followed by the appearance of a large variety of clones with an enormous diversity of HLA expression.¹³ T cell-mediated immunoselection of HLA-I negative tumour cells is taking place, ending with the tumour nest being uniformly HLA-I negative.¹⁴ We have named these two phases as ‘permissive’, when T cells are entering the tumour, and ‘encapsulated’ when the tumour is uniformly HLA-I negative, with T cells excluded to the stroma surrounding the tumour nest⁵ (Fig. 1A).

Fig. 1.

Examples of MSS and MSI-H HLA-I negative CRC tumours. Cryopreserved tissue sections of CRC tumours stained with the anti-HLA-I mAb w6/32 (brown staining). A IHC staining of an HLA-I negative MSS CRC tumour in which infiltrating cells are absent in the tumour nests (black arrow) and are located in the tumour stroma (red arrow) that is HLA-I positive. B IHC staining of an HLA-I negative MSI-H CRC tumour where infiltrating cells are located inside the tumour nests (black arrow). The stroma is HLA-I positive (red arrow)

Loss of HLA-I on tumour cells represents a serious obstacle in the development of effective anti-tumour immunotherapies since it renders the T cells targetless.⁷ In this sense, it was reported that melanomas with defects in antigen presentation were resistant to immune checkpoint inhibition (ICI).¹⁵ In a colorectal cancer (CRC) case study, adoptive transfer of ex vivo expanded tumour-infiltrating lymphocytes (TILs), specific for a mutant KRAS G12D epitope presented on HLA-C*08:02, resulted in the loss of this allele and progression of the tumour.¹⁶ These data suggest that the loss of HLA-I can result in both initial resistance and secondary immune escape after immunotherapy. Understanding the molecular mechanisms behind the loss of HLA-I is crucial in the design of therapeutic targeting with the aim of restoring HLA-I expression.^17–19

This review will discuss the different altered HLA-I phenotypes observed during the natural history of CRC development with special emphasis dedicated to define the different molecular mechanisms responsible for the loss of HLA-I. This important biological phenomenon will also be analyzed in the context of the recently identified consensus molecular CRC subtypes (CMSs) and its important implications in tumour spread and resistance to ICIs. Finally, we will consider alternatives for therapeutic intervention to overcome/circumvent HLA-I-mediated immune escape and also discuss the expression of HLA-II molecules in CRC.

Altered HLA-I expression in CRC

The absence of HLA-I in CRC was described a long time ago by different groups, including ourselves.^20–27 It is known that the transition from a normal colorectal epithelium to adenoma and later to carcinoma is associated with the acquisition of different cumulative molecular events such as loss of chromosomes 5q, 17p and 18q and mutations in the KRAS oncogene.^28,29 It was also clearly shown that the normal epithelia, as well as benign and premalignant lesions of the colon, are HLA-I positive with no apparent alterations.^1,30 For instance, colon adenomas that are known to accumulate several activated protooncogenes such as KRAS are HLA-I positive.²³ However, recent molecular studies have shown that monoallelic genomic lesions and mutations in the HLA-I and β2-microglobulin (β2m) loci can be detected in HLA-I positive CRC adenomas.^31,32 In addition, rare focal losses of certain HLA-A and -B alleles have been observed in intestinal adenomas by immunohistochemistry (IHC).^30,33,34 These data suggest that while homogeneous loss of HLA-I cell surface expression is not detected in CRC premalignant lesions, the development of immune escape variants starts early on during CRC progression, similar to that seen in lung cancer.³⁵

Early studies in cryopreserved CRC tumour tissues using anti-HLA mAbs, directed against monomorphic HLA determinants, revealed HLA-I losses in 30–40% of the tumours.³⁶ However, when a broader panel of mAbs against monomorphic, locus-specific and allele-specific determinants was used, the incidence of HLA losses increased up to 73%.³⁷ Similar altered HLA-I phenotypes were identified in CRC cell lines using a variety of techniques.^38,39 Despite the high frequency of altered HLA expression detected in these tumours, it was not possible at that time to study all the HLA alleles suggesting that HLA-I losses were underestimated⁴⁰ (Fig. 2).

Fig. 2.

CRC developmental stages and HLA-I expression. The progression from normal colonic epithelium to malignant CRC follows two main pathways. Methylation of MLH1 results in DNA mismatch repair (MMR)-deficiency and microsatellite instability (MSI-H) and progress through mutations in BRAF, PTEN and TGFβRII in a CIMP setting. In contrast, microsatellite-stable (MSS) tumours suffer an initial inactivation of the adenomatous polyposis coli (APC) gene followed by KRAS mutations and later SMAD4 and TP53 inactivation. The normal intestinal epithelium is HLA-I positive. MSI-H and MSS adenomas are HLA-I positive but start to accumulate mutations and loss of heterozygosity (LOH) in the HLA-I and β2m loci. Finally, a strong immune pressure in CRC tumours results in total HLA-I loss and immune escape

CRC is a heterogeneous disease that can be divided in two major groups.⁴¹ First, 15% of CRC tumours are ‘hypermutated with microsatellite instability (MSI-H)’ because of a defect in the DNA mismatch repair (MMR) system and ubiquitous somatic mutations in repeated DNA sequences that occur during colon carcinogenesis.⁴² Sporadic MSI-H CRC exhibit a CpG island methylator phenotype (CIMP), resulting in methylation of MLH1, whereas hereditary nonpolyposis colorectal cancer, or Lynch syndrome, is due to germline inactivation of MMR genes.^32,43 Second, 85% of CRC cases belong to the ‘non-mutated, microsatellite stable (MSS) group’.⁴⁴ These two groups of CRC also have different molecular mechanism responsible for their HLA-I alterations.⁴⁵ Recently, Guinney et al. performed a large-scale analysis of six previously published bulk-transcriptome data sets from CRC tumours (stage I–IV) and proposed four CMSs with distinct molecular and clinical features, encompassing >85% of all tumours. In this study, the MSI-H tumours were grouped into CMS1, while MSS tumours were divided into CMS2, CMS3 and CMS4 tumours.⁴⁶ Below we will discuss in more details the prevalence and mechanisms behind HLA-I loss in MSI-H and MSS CRC tumours, taking into consideration their new molecular classification into CMS1–4.

HLA-I loss in MSI-H CRC

HLA-I loss in MSI-H tumours depends primarily on structural/irreversible ‘hard lesions’ due to a lack of β2m synthesis or to the synthesis of a truncated β2m caused by mutations. A total loss of HLA-I has been seen in up to 70% of MSI-H CRC tumours.^32,47–50 Loss of β2m is gradually increasing with tumour stage I-III but is rarely seen in metastatic disease.³² This apparently contradictory observation and the underlying mechanisms will be discussed below in the section concerning HLA-I expression in CRC metastases. In addition to β2m mutations, further silencing mutations have been detected in the HLA-A, -B and -C genes and in genes important for the expression of HLA-I (NLRC5) and antigen-presenting machinery (APM) genes, including TAP-1, TAP-2, tapasin, Erp57, calreticulin and calnexin.^50–55 Although MSI-H CRC tumours exhibit low frequency of somatic copy number alterations (SCNAs), several reports have demonstrated loss of heterozygosity (LOH) at the β2m and HLA-I loci.^52,53,56 Grasso et al. found that LOH can occur at sites of disruptive mutations in β2m, resulting in biallelic β2m inactivation.⁵³ Importantly, the impact of such a lesion on the efficacy of ICI therapy was recently demonstrated in a case study where a frameshift mutation, followed by LOH, of the β2m locus rendered a MSI-H CRC tumour resistant to anti-PD-1 immune therapy.⁵⁶ Lynch syndrome is a hereditary CRC due to germline mutations in hMLH1, hMSH2, hMSH6 and PMS2 resulting in hypermutated MSI-H tumours, making up 2–4% of all CRC cases.⁵⁷ Lynch syndrome tumours exhibit HLA-I loss in up to 50% of the cases which is due to a total loss of β2m in half of the cases.^32,52,58,59

In most tumours, HLA-I negativity is associated with the exclusion of infiltrating mononuclear cells that remain in the stroma outside of the tumour nest.⁵ Interestingly, HLA-I negative MSI-H CRC tumours are heavily infiltrated tumours, representing an ‘exception to the rule’. This is visualized in Fig. 1A, B, which show examples of HLA-I negative MSS and MSI-H cryopreserved tumour tissues, respectively.

According to the CMS-stratification of CRC tumours, the majority of MSI-H tumours are grouped into CMS1 due to their strong enrichment of genes associated with immune infiltration and activation of immune evasion pathways, including PD-1.⁴⁶ Several studies have shown that CMS1 tumours exhibit high levels of HLA-I/β2m mRNA expression despite the high degree of HLA-I loss observed at the tumour cell surface.^60–62 This finding could be explained by the high numbers of HLA-I positive infiltrating cells/TILs found in MSI-H/CMS1 tumours and the observation that HLA-I/β2m mRNA and protein expression do not always correlate.^45,47 These data highlight the difficulties in interpreting HLA-I expression on tumour cells using RNA-expression data.

HLA-I loss in MSS CRC

MSS tumours represent the vast majority of CRC cases (85%). Different altered HLA-l phenotypes (total loss, haplotype loss, allelic loss) can be found in up to 75% of MSS tumour samples,^37–40 while total HLA-I loss is seen in 30–40% of the cases.^50,52,63 In contrast to MSI-H tumours, HLA-I loss in MSS tumours depends primarily on transcriptional downregulation of the TAP-1/TAP-2 and LMP7 genes.^45,54 MSS CRC features a high frequency of SCNAs due to defects in chromosomal segregation and DNA repair.⁶⁴ Consequently, HLA-I haplotype loss is a common mechanism that generate escape variants, and can be observed in up to 40% of MSS CRC tumours.^52,65,66 Thus, the molecular mechanisms responsible for HLA-I loss in MSS CRC result in most cases from a combination of an HLA haplotype loss due to chromosome 6 total or partial loss,⁶⁷ associated with a coordinate transcriptional downregulation of HLA-I heavy chain, β2m and APM genes⁴⁵ (Fig. 3).

Fig. 3.

HLA-I loss mechanisms in CRC tumours. The molecular mechanisms responsible for HLA-I loss are subdivided in reversible/‘soft’ and irreversible/‘hard’ lesions. This classification has important clinical implications since the recovery of ‘soft’ lesion can be achieved by conventional or new immunotherapy such as immune checkpoint antibodies. The local release of TH1 cytokines in the tumour microenvironment can recover HLA-I expression and induce metastatic regression. The ‘hard’ lesions are due to mutations or chromosome loss and do not recover HLA-I. These metastatic lesions are resistant to immunotherapy and progress. APM (antigen processing machinery), LOH (loss of heterozygosity), TH1 (T-helper type 1)

Recently, MSS tumours have been divided into CMS2-4 tumours based on specific gene expression signatures related, in part, to WNT/MYC signalling, KRAS mutations and TGF-β signalling.⁴⁶ Analysis of HLA-I expression in CMS2-4 tumours by IHC suggests that HLA-I loss/downregulation can be found in all CMSs but is more prevalent in CMS4 in comparison to CMS2/3 tumours.^63,68 We will briefly mention which oncogenic pathways are enriched in each MSS CMS subgroup and how they might contribute to the HLA-I-mediated immune escape.

Firstly, CMS2 tumours represent approximately 50% of all MSS CRC cases, and exhibit a marked increase in WNT and MYC signalling, which correlates with immune exclusion in many cancers, including CRC.^53,69,70 WNT/β-catenin/MYC signalling has been shown to repress the expression of HLA-I and APM genes in glioma and melanoma cell lines,^71,72 suggesting that this pathway also could contribute to HLA-I downregulation in MSS CRC tumours.

Secondly, CMS3 tumours are highly enriched in oncogenic KRAS mutations which results in a constitutive activation of the MAPK (Raf/MEK/ERK) pathway^60,73. Oncogenic KRAS has been found to correlate with lower HLA-I and APM (TAP-1, LMP2, LMP7) expression in CRC tumours.^74,75 Several studies using human CRC cell lines or employing animal models of CRC suggest that oncogenic KRAS can inhibit HLA-I expression, in part by suppressing IFN-γ signalling (STAT1, STAT2, IRF2).^76–78 These data indicate that aberrant MAPK signalling could participate in driving immune evasion and resistance to ICIs in CRC by reducing HLA-I expression.

Finally, CMS4 tumours are characterized by prominent stromal invasion, TGF-β activation, angiogenesis, and contain high levels of M2 macrophages and regulatory T cells. CMS4 tumours are significantly associated with advanced disease (stage III–IV) and have a poor prognosis.^46,79,80 Interestingly, we have recently shown that the tumour tissue architecture changes with HLA-I loss. In HLA-I positive tumours, FAP-1+ cancer-associated fibroblasts (CAFs) are mixed with tumour cells, lymphocytes and M1 macrophages, forming a loose network without clearly defined tumour nests. In contrast, HLA-I negative tumours are encapsulated by a dense network of FAP-1+ CAFs, containing anti-inflammatory M2 macrophages, resembling a T-helper type 2 (TH2)-like granuloma response with CD8+T cells restricted to the peritumoral stroma (see Fig. 1A).^5,81,82 Importantly, such tumour tissue organization has been observed in a variety of tumours of different histological type, including CRC; however, it has not previously been associated with the absence of HLA-I expression.⁸¹ Little is known about the mechanisms behind the remodelling of the tumour stroma in MSS HLA-I negative CRC tumours and its consequence for treatment resistance. Our group found an upregulation of TGF-β1 and -β3 in HLA-I negative MSS CRC tumours,⁴⁷ suggesting that these tumours could belong to the CMS4 subtype. Early studies in mouse showed that TGF-β1-deficient mice exhibit a dysregulated overexpression of HLA-I and HLA-II, especially in heart, lung and kidney, independently of IFN-γ expression. Furthermore, TGF-β1 can inhibit the IFN-γ-mediated induction of HLA-I on mouse embryonic fibroblasts and uveal melanoma cells.^83,84

Recently, we have tried to determine the CMS identity of HLA-I negative MSS CRC tumours using the CMS classifier software.⁴⁶ The obtained results indicate that MSS HLA-I negative CRC tumours do not classify only as CMS4 but relates to both CMS2 and CMS4 (unpublished observations). In addition, the MSS HLA-I negative tumours might also belong to the recently identified ‘Wound Healing’ immune subtype characterized by angiogenesis, extracellular matrix remodelling and TH2 responses.⁸⁵ This phenotype shares common characteristics with our previously reported ‘encapsulated’ group observed in the late phase of T infiltration into the tumour nest. Understanding the nature and consequence of the stromal rearrangements in the HLA-I negative tumours could aid in the development of therapies that reverse the suppression of HLA-I and improve the efficacy of ICIs. We favour the idea that the stromal rearrangement associated with the tumour capsule is a consequence of the end of the T cell-killing of HLA-I positive tumour cells and the beginning of a new phase characterized by a TH2-type response.^5,81 Thus, CMS2, -3 and -4 could represent different stages of the MSS CRC tumour progression as we already reported for lung, bladder, prostate and melanomas.⁸¹

HLA-I expression in colorectal cancer metastasis

Cancer immunotherapy is implemented when patients are suffering from metastatic disease, months or years after the primary tumour lesion was diagnosed and surgically removed. It is therefore relevant to know the HLA-I expression in these lesions and the molecular mechanisms responsible for the HLA-I loss or downregulation. We have shown in mouse fibrosarcomas and human tumours that the primary tumour lesion is highly heterogeneous for H-2 and HLA-I expression with multiple tumour clones expressing different amount of HLA-I.^5,86 However, in metastatic lesions the HLA-I expression (positive, intermediate or negative) is homogeneous, suggesting that the lesion derives from a single tumour clone from the primary tumour lesion¹⁴ (Fig. 4).

Fig. 4.

HLA-I heterogeneity in primary tumour lesions versus homogeneous metastases. Primary tumours are heterogeneous, composed of different clones with different levels of HLA-I expression. Immune selection produces homogeneous metastatic lesions in different organs that can be HLA-I positive, -negative or with intermediate levels of HLA-I expression

There are reports in different tumours, in particular in melanoma and CRC, showing that metastatic colonies can have different HLA-I phenotypes.^27,32,87,88 The route used for metastatic spread plays an important role in selecting the final HLA phenotype of the metastasis. For instance, it is well established that liver metastases originating from β2m-mutated MSI-H CRC tumours are rare and HLA-I positive.^32,89 The selection of HLA-I positive blood-born liver metastases derived from MSI-H CRC tumours could represent an example of the NK cell-mediated elimination of metastatic cells that lack HLA-I expression.⁹⁰ In this context, it has been reported that partial downregulation of HLA-I in CRC has worse prognosis than complete HLA-I loss.⁹¹ These observations indicate an important role for NK cells in eliminating HLA-I negative metastatic clones and selecting HLA-I positive metastatic cells during dissemination via the blood. In this context, it is well known that HLA-B alleles carrying the Bw4 epitope can interact with NK inhibitory receptors, controlling NK cell cytotoxicity. In one study, HLA-I expression at the allelic level was compared between leukaemia cells and autologous normal cells and the downregulation of HLA-A and/or HLA-B allospecificities was detected in the majority of patients. Interestingly, this downregulation affected only HLA alleles of the HLA-Bw6 subgroup that do not interact with NK cell receptors allowing leukaemic cells to inhibit NK cell and escape killing.⁹² This example indicates that the selective downregulation of HLA-A and HLA-Bw6 allospecificities, while preserving HLA-Bw4, provides leukaemic cells with an escape mechanism not only from cytotoxic T lymphocytes (CTLs) but also from NK cells.⁹⁰ However, there are indications that metastases derived from MSI-H CRC cells travelling through lymph nodes to different locations can be HLA-I negative, as shown by a recent study using paraffin-embedded CRC metastases obtained from various organs, suggesting again the importance of the route used for metastatic spread.⁶³ It is also possible that the altered metastatic HLA phenotype is ‘new’, i.e., acquired during the metastatic migration and absent in the primary tumour. This might represent a new round of ‘Generation of Diversity’ and Darwinian type of T-/NK cell-mediated immune selection.¹⁴ In this context, we recently studied HLA-I expression by flow cytometry, in a panel of CRC cell lines derived from MSS and MSI-H CRC, both from primary tumours and autologous metastases (provided by Dr. Michael Linnebacher, University of Rostock, Germany). We observed that HLA-I expression was weakly positive in all cell lines, with poor responses to IFN-γ. Noteworthy, cell lines derived from liver metastases were HLA-I positive (unpublished data).

A relevant factor for the prediction of responses to immunotherapy is the nature of the molecular defect of HLA-I (reversible or genetic) in the metastatic lesion. We believe that it is necessary to carefully monitor the HLA expression in metastatic tumour lesions using a variety of techniques. The complexity of the HLA system requires a well trained personnel in order to perform an accurate diagnose of the HLA profile in tumour tissues.⁹³

Strategies to overcome immune escape due to HLA-I loss

The recovery of HLA-I antigens is a major future challenge since it will allow memory T lymphocytes to again recognize the tumour antigens they were sensitised during tumour progression, before the escape took place.

The restoration of tumour HLA-I expression is particularly important for boosting the efficacy of the existing protocols of cancer immunotherapy, most of which aim at stimulating T cell-mediated anti-tumour cytotoxic responses. There is accumulating evidence suggesting that immunotherapy is effective in eliminating HLA-I positive tumour cells, while cells with loss or downregulation of HLA-I escape the therapy-induced immune attack and produce new distant tumour lesions. The important determining factor of the clinical response to cancer immunotherapy is the molecular mechanisms underlying the HLA-I alterations in cancer cells. Thus, HLA-I downregulation produced by reversible/‘soft’ molecular alterations can be corrected in vitro by IFN-γ or other cytokines^17,94 or recovered by immunotherapy that can stimulate a release of T-helper type 1 (TH1) cytokines in the tumour microenvironment.⁸⁸ As a result, primed T cells can again recognize tumour antigens and induce cancer regression. In contrast, immunomodulatory interventions cannot correct structural or irreversible/‘hard’ alterations caused by mutational events and chromosomal abnormalities, and it may lead to the progression of HLA-I negative lesions.

Correction of ‘HARD’ molecular lesions

There is accumulating evidence indicating that resistance to ICIs is frequently associated with genetic aberrations in HLA-I and β2m genes, and the IFN signalling pathway.^12,15,56 In melanoma patients, resistance to anti-PD-1 therapy was reported to be linked to structural alterations and reduced expression of HLA-I, APM and/or IFN signalling components.^12,15 Since primary tumours mostly display a heterogeneous HLA-I expression with both types of alterations, immunotherapy sometimes produces mixed responses by generating both regressing and progressing metastases in the same patient.

We observed that in two melanoma patients, immunotherapy with interferon-α-2b or autologous vaccination plus BCG (M-VAX vaccine) induced regression of some metastases (showing high HLA-I expression by IHC), and progression of other lesions with low HLA-I expression due to LOH at chromosome 15 and a reduced mRNA expression of HLA-I genes.⁸⁸ Similarly, we reported that BCG immunotherapy in superficial non-invasive bladder cancer induces primary tumour rejection. However, recurrent tumours had more profound alterations in HLA-I with higher incidence of LOH at chromosomes 6 and 15.⁸

Structural genomic lesions resulting in the loss of β2m or heavy chain and APM genes can be counteracted with gene therapy. We have reported the recovery of HLA-I expression in β2m-deficient melanoma and CRC tumour cell lines following infection with adenovirus encoding the β2m gene. Restoration of β2m resulted in HLA-I reexpression, followed by HLA-I-mediated recognition and destruction of the tumour cells by CTLs.^95,96 We believe that local transfer of the wild type HLA-I gene/s into the metastatic lesion could recover the altered antigen presentation and T cell-mediated rejection.⁹⁷ This strategy may provide an additional tool at advanced stages when HLA-I negative cells dominate distant metastatic lesions in many types of cancer. CRC metastases that lose β2m expression and do not respond to immunotherapy could also potentially benefit from this approach to recover antigen presentation.⁵⁶ Gene therapy strategies can also be used to transfer genes coding for potential regulators of HLA and APM gene transcription, including FHIT and NLRC5. In a mouse model of fibrosarcoma, transfection of H-2 negative tumour cells with the FHIT tumour suppressor gene restored HLA-I expression, resulting in tumour rejection by T cells.⁹⁸ Similarly, transfection of the poorly immunogenic B6 melanoma cell line with NLRC5 increased its H-2-mediated presentation of tumour antigens to CD8+ T cells and significantly reduced tumour growth in vivo.⁹⁹ Interestingly, Kalbasi et al. demonstrated the induction MHC class I in JAK1-deficient murine and human melanoma cell lines, independently of IFN signalling and NLRC5.¹⁰⁰ In summary, gene therapy and alternative approaches to induce HLA-I represent promising treatment modalities to overcome tumour immune escape.

Correction of ‘SOFT’ molecular lesions

Dysfunctional transcriptional, post-transcriptional and translational control of antigen presentation are important reversible mechanisms of HLA downregulation. Tumour cells with these ‘soft’ lesions have better chances to upregulate HLA and therefore to be eradicated by T lymphocytes activated by immunotherapy-induced cytokines, such as IFN-γ.

IFN-γ plays an important role in the anti-tumour immune response through its transcriptional regulation of HLA-I/II, β2m and several APM components.¹⁰¹ Epigenetic silencing of interferon-inducible genes in the JAK/STAT pathway can result in primary and secondary immune escape in melanoma.^{12,94,102,103} In CRC, high STAT1 activation levels in the tumour is an independent predictive factor for good prognosis and its expression correlates with a high levels of TILs.¹⁰⁴ Importantly, low-dose IFN-γ administration to patients with metastatic melanoma resulted in an increase in HLA-I and HLA-II expression on the tumour cells.¹⁰⁵ In a recent study, patients with immunologically cold sarcomas were treated with systemic IFN-γ injections that resulted in an increase in HLA-I expression and infiltration of CD8+ TILs.¹⁰⁶

Another attractive approach for the restoration of HLA-I expression due to ‘soft’ lesions is the use of epigenetic modifiers, such as inhibitors of DNA methyltransferases or histone deacetylases (HDAC). We have shown in a melanoma cell line that HLA-I expression and antigen-specific CTL responses can be restored after 5-aza-2’-deoxycytidine treatment.¹⁰⁷ Demethylation of HLA-I and APM-related genes, including the TAP and LMP2 genes,^108,109 increased their sensitivity to IFN-γ-mediated induction of HLA-I. In breast cancer and CRC patients, the DNA methyltransferase inhibitor 5′-azacytidine in combination with the HDAC inhibitor entinostat resulted in the upregulation of tumour HLA-I.^110,111 Recently, HDAC inhibitors have been demonstrated to sensitize murine B cell lymphomas to PD-1 blockade by upregulating their expression of MHC class I.¹¹²

One of the leading modern cancer therapy approaches represents targeting tumour cell signalling pathways, including MAPK inhibition. As discussed above, the aberrant activation of the MAPK pathway can suppress the expression of HLA-I and HLA-II, resulting in immune escape. Systemic inhibition of MAPK kinase (MEK) has been shown to increase HLA-I expression and immune infiltration in several murine models of tumour growth in vivo. Although MEK inhibition alone reduced tumour growth, the simultaneous blockage of PD-L1 was needed to achieve a durable regression.^113–115 Several clinical trials are currently analyzing the effects of MEK inhibitors in combination with ICIs for the treatment of MSS CRC.¹¹⁶

In another study, the anti-EGFR blocking antibodies cetuximab and nimotuzumab demonstrated increased tumour HLA-I expression, as a result of augmented APM mRNA levels, and susceptibility to CD8+ T cell-mediated lysis.¹¹⁷ In this context, in vitro treatment of thyroid cancer cells with selumetinib, in combination with interferon, upregulated HLA-I expression.¹¹⁸

Lastly, there are multiple ongoing studies exploring the role of immunotherapy in conjunction with low-dose radiation and chemotherapy as well as initiating immunotherapy at earlier stages of malignancy.¹¹⁹ In animal models and in vitro experiments, it has been demonstrated that low doses of radiation and chemotherapy enhance antigen presentation and tumour elimination by T cells, and is linked to the production of neoantigens, their degradation and release of novel peptides.^120,121 Reits et al. showed that immunotherapy of murine colon adenocarcinomas was successful only when preceded by radiotherapy.¹²²

In summary, upregulating HLA-I expression on tumour cells harbouring ‘soft’ molecular lesions can be achieved by TH1-type cytokines, HDAC inhibitors, inhibition of MAPK signalling and by ICI alone or in combination with radiation and chemotherapy. A combination of different strategies could lead to the upregulation of HLA-I expression and activation of anti-tumour T lymphocyte responses.^18,123

Beyond all these strategies, other ways to overcome cancer immune escape associated with HLA-I loss are under active investigation, including activation and redirection of NK cells,^124,125 the use of TEIPP-specific CD8+ T cells^126,127 and virtual memory CD8+ T cells that has been shown to destroy chemotherapy-treated MHC class I-negative tumours.¹²⁸

HLA class II expression in colorectal cancer

According to classical immunology, constitutive HLA-II expression is restricted to the professional antigen-presenting cells (dendritic cells, macrophages and B cells). However, it has been demonstrated that HLA-II can be expressed by epithelial cells under physiological conditions¹²⁹ or become induced in certain situations by proinflammatory cytokines, including IFN-γ.^130–133

It was reported years ago that HLA-II can be expressed on tumours of different origin, including laryngeal, breast, gastric, lung, lymphomas, melanomas and CRC, and in most cases this expression is associated with a better prognosis and overall survival time.^{22,23,133,134} The HLA-II molecules are required for tumour-associated antigen recognition by CD4+ T cells and it has been hypothesized that their expression by tumour cells may reflect an effective antigen presentation to T cells and a cross-presentation of tumour-associated antigens.¹³⁵

HLA-DR is the predominant HLA-II molecule expressed in CRC tumours and its expression has been associated with the intensity of T cell infiltration and correlates with a better prognosis.^131,136,137 The normal intestinal mucosa does not express HLA-II but its expression is frequently detected in mucosa adjacent to CRC tumours.^23,138 Furthermore, the intensity of HLA-II expression in adenomatous polyps has been shown to increase with the degree of dysplasia.²³

Today we know that the expression of HLA-II in CRC, in particular HLA-DR, is different in different CRC subgroups.^136,139 A higher frequency of HLA-DR-expressing tumour cells has been observed in MSI-H CRC, in comparison to MSS CRC.^131,137,138 These data suggest that apart from methodological differences in staining and evaluating HLA-DR positivity (focal/patchy staining pattern versus homogeneous pattern), the degree of HLA-DR expression in CRC could be determined by the CRC subtype.

In this context, we have recently observed that in MSS and MSI-H tumours in Lynch syndrome patients, the frequency of HLA-DR is around 12%. In contrast, this percentage is notably increased in sporadic MSI-H patients (63%) (unpublished data). It is known that IFN-γ is highly upregulated in CRC tumours with a dense lymphocytic infiltration, which correlates with an induction of HLA-DR.^137,140 However, our data show that both MSI-H Lynch syndrome and sporadic MSI-H tumours are heavily infiltrated by lymphocytes, resulting in no correlation in both groups between HLA-DR expression and the degree of infiltration. These data suggest that the high HLA-DR expression pattern observed in sporadic MSI-H tumours might be constitutive and not associated with local cytokine release by infiltrating cells. The mechanisms behind the constitutive expression of HLA-DR in sporadic MSI-H tumours are not clear but could be due to their CIMP-high phenotype (not present in Lynch syndrome CRC) that can regulate HLA-DR expression.¹⁴¹

In some sporadic MSI-H tumours, we observed a patchy HLA-II expression pattern. In some tumour areas, HLA-DR-positive tumour cells were surrounded by a HLA-DR-negative peritumoral stroma and vice versa (HLA-DR-positive stroma adjacent to HLA-DR-negative tumour), which suggests a different grade of lymphocyte activation in these areas (Fig. 5A, B). These data indicate that loss of HLA-II expression in sporadic MSI-H CRC could be due to a strong immunoselective pressure, as reported for HLA-I.^53,138 Loss of HLA-II in sporadic MSI-H tumours could be the consequence of inactivating mutations in genes that regulate HLA-II expression, including RFX5, CIITA, and RFXAP.¹⁴² Mutations, but not epigenetic silencing, in these genes are frequent in MSI-H sporadic tumours and are associated with a high infiltration of CD4+ T cells into the tumours.^138,139 This apparent contradiction can be explained by a strong immunoselection of HLA-II-mutated clones mediated by the high infiltration of CD4+ T cells (Fig. 5B). In fact, Grasso et al.⁵³ established four MSI-H CRC clusters based on HLA-I and II expression, being cluster II, of HLA-I/HLA-II-negative phenotype, in which more samples with mutations for NLRC5/RFX5 were found. This is in agreement with the role of RFX5 as an essential regulator of HLA-II expression, thus being able to promote immune evasion in MSI-H CRCs.

Fig. 5.

An example of heterogenous HLA-II (HLA-DR) expression in a sporadic MSI-H CRC tumour. A A sporadic MSI-H CRC paraffin-embedded tumour tissue was stained with an anti-HLA-DR mAb. The tumour is HLA-DR positive (yellow arrow), whereas the infiltrating cells are HLA-DR negative (red arrow). B A different area of the same MSI-H tumour tissue shows complete loss of tumour HLA-DR expression (yellow arrow), while the cellular infiltrate is now HLA-DR positive (red arrow). The degree of infiltration is similar in the HLA-DR positive and negative areas of the tumour samples demonstrated in A and B

In summary, we propose that the expression of HLA-DR in sporadic MSI-H CRC is constitutive due to unknown epigenetic mechanisms. HLA-II loss in this subgroup is probably associated with hard lesions in genes regulating HLA-DR expression. In contrast, MSS and hereditary MSI-H tumours express lower levels of HLA-DR that can be upregulated by the release of IFN-γ in the tumour microenviroment by infiltrating mononuclear cells.

Conclusions

A revival of cancer immunotherapy is currently taking place due to the strong clinical responses observed in some patients after treatment with ICIs, especially in patients with metastatic progression. However, not all patients respond, and not all lesions regress. The reason for such different responses is still unknown. We have proposed that the loss or downregulation of HLA expression in primary tumours and metastatic lesions play an important role in resistance or secondary immune escape after ICI therapy, due to their well-established role in tumour antigen presentation to T lymphocytes and to the modulation of NK cell activity. This review focuses on immune evasion in CRC, which is an important and rapidly evolving topic. CRC is a heterogeneous disease with different molecular pathways of carcinogenesis and various patterns and underlying mechanisms of tumour HLA alterations. In all, 15% of CRC belong to the MSI-H group and 85% to the MSS group. These two CRC tumour types represent distinct entities with different pathways inducing HLA-I altered phenotypes. The MSS group behave as other tumours we have studied (lung, breast, bladder cancers or melanoma), showing a high incidence of LOH at chromosome 6 and 15 as a major molecular mechanism producing tumours harbouring HLA haplotype losses.^67,143 MSS CRC tumours also lose HLA-I expression via a coordinated transcriptional downregulation of the HLA-I heavy chain, β2m and APM genes. The infiltration pattern in this MSS group is associated with the expression of tumour HLA molecules, i.e., a higher HLA expression on the cancer cells correlates with an increased number of TILs, and vice versa. This ‘permissive’ phase later evolves into a tumour composed only of HLA-I negative tumour cells with lymphocytes located in the stroma, a so-called encapsulated phase. We favour the idea that the new consensus molecular subtypes CMS2, -3 and -4, mentioned in this review, could represent different stages of the MSS CRC evolution during the natural history of a particular tumour in a given patient. In contrast, the MSI-H subgroup, representing CMS1 tumours, behave in a very peculiar way by generating HLA-I losses through an accumulation of mutations in the HLA and β2m genes with a characteristic pattern of T cell infiltration. In this review, we have discussed these differences in detail, also making an emphasis on the HLA-II expression patterns found in MSS, sporadic MSI-H and Lynch syndrome tumours. We believe that the analysis of HLA expression in metastases, together with the identification of the responsible molecular mechanism, provides important information required to identify patients and/or malignant lesions that may respond to immunotherapy.¹⁴ There is still a long way to go, but there is no doubt that in the near future, tumour HLA-typing will become a clinical routine, as it has been in case of peripheral blood for organ transplantation.¹⁴³ Finally, we would like to emphasize that the HLA-I losses frequently observed in human tumours of different origin should not be seen as an obstacle for a successful T cell-mediated immunotherapy, but as a crucial step during the natural history of tumour development.

Competing interests

The authors declare no competing interests.