Lymphocyte activation gene 3 (LAG3) protein expression on tumor-infiltrating lymphocytes in aggressive and TP53-mutated salivary gland carcinomas

Setting boundaries to T cell activity through inhibitory co-receptors (IR) is part of the physiological regulation of immune responses. Conversely, in the setting of a sustained and uncontrolled local presentation of neoantigens by tumor cells (TC), the upregulation of IRs can lead to T cell exhaustion in the tumor micro-environment (TME) [1]. Blockage of these inhibitory stimuli is the principle of checkpoint inhibitors. Most prominently among this group, inhibitors aiming at programmed cell death protein 1 and its ligand 1 (PD-1/PD-L1) have recently led to a breakthrough in targeted cancer therapy and are nowadays used for the treatment for a multitude of neoplastic diseases, including nonsmall cell lung cancer (NSCLC), melanoma and urothelial carcinoma [2]. Despite the success of a PD-1/PD-L1 blockage in some patients, the majority of patients do not respond [3]. This led to the therapeutic exploration of alternative IRs, including lymphocyte activation protein 3 (LAG3).

LAG3, a co-receptor of the T cell receptor, is structurally highly homologous to CD4 and shares its main ligand MHC class II, binding to it with a much higher affinity than CD4 [4]. In melanomas and colorectal cancer, LAG3 is predominantly expressed on FOXP3+ regulatory T cells that exert an immunosuppressive function through release of TGF-ß and IL10 in the TME [5]. A co-expression of LAG3 and PD-L1 on tumor-infiltrating lymphocytes (TILs) has been observed in several tumor types [6]. Dual blockage of LAG3 and PD-L1 with monoclonal antibodies led to increased CD4- and CD8-positive TILs and tumor clearance compared to monotherapy in mice models [7]. Head and neck squamous cell carcinoma (HNSCC) and colorectal carcinoma with LAG3-expressing TILs were associated with lymph node metastasis, larger tumor size and shorter overall survival in the case of nodal status-negative HNSCC [8].

At present, numerous clinical trials investigate the efficacy of LAG3 blockage, enrolling patients with a broad spectrum of neoplastic diseases (e.g., head and neck carcinomas, melanoma, colorectal carcinoma, NSCLC: NCT01968109).

The relatively rare carcinomas of the salivary glands (SGC) make up fewer than 0.5% of all cancers and can be subdivided into distinct histologic types, with mucoepidermoid carcinoma (MEC) as the most frequent subgroup. Entity-specific genomic translocations with an oncogenic driver potential were revealed for adenoid cystic carcinomas (AdCys), MEC and more recently, secretory carcinomas and acinic cell carcinomas [9]. Currently, with the exception of secretory carcinomas, these genomic alterations cannot yet be therapeutically exploited [10]. The biological aggressiveness is overall moderate compared to other carcinomas, but it varies significantly across different subtypes. For instance, the prognosis of salivary duct carcinomas (SDC) is poor (mean overall survival: 56 months) [11]. Therapy consists of complete excision, followed by stage- and risk-adapted radiotherapy. In a setting of irresectability, nodal or distant metastasis, prognosis is unfavorable (67% and 34% 5-year OS, for regional and distant metastasis, respectively) and only untargeted chemotherapy with short response duration is at hand for most entities as first-line treatment. The demand for therapeutic targets prompted the exploration of inhibitory immune-checkpoint molecules expression in SGC: Recent studies presented divergent results for PD-L1 expression on tumor cells and TILs [12‐16]. PD-L1 positivity was associated with poor outcome and higher UICC stage (Union International Contre le Cancer) [12, 17]. Also, programmed death ligand 2 (PD-L2) and cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) were expressed on SGC tumor cells at frequencies of 63% and 74.1%, respectively, whereby PD-L2 expression was associated with a shorter progression-free survival [13]. The first clinical study with immune-checkpoint inhibitors in SGC applied the PD-1 inhibitor pembrolizumab in a metastatic/recurrent disease setting. A relatively high rate of responders (12%) and a manageable safety profile raised hopes for immunotherapeutic options in SGC [18]. Another comparable phase II study reported a slightly higher response rate of 16% for SGC (only partial responses) [19]. Also recently, a phase II trial demonstrated encouraging effectiveness of the PD-1 inhibitor nivolumab in combination with CTLA-4 inhibitor ipilimumab in SGC [20]. Another study which evaluates the therapeutic use of nivolumab alone is ongoing (NCT03132038). While a series of studies investigated the role of PD-L1, PD-L2 and CTLA-4, little is known about the presence of other IRs in SGC.

Somatic TP53 mutations are considered to be the most frequent genetic alterations in cancer cells of different origin. TP53 immunohistochemistry (IHC) is able to detect and discriminate between nonsense (indel, stopgain and alternative splicing) and missense (nonsynonymous) mutations: While the former lead to a complete loss of TP53 protein expression, the latter usually result in nuclear TP53 protein overexpression [21]. Missense mutations occur more often (87.9%) than nonsense or silent mutations, but either can lead to a loss of function. Intact TP53 acts as a tumor suppressor gene, initiating cell cycle arrest and apoptosis under genotoxic stress. A deregulated proliferational capacity due to a loss of TP53 function seems to be the main selective advantage for TP53 mutations [22] and allows for a higher mutational burden [23]. Consecutively, a higher number of neoantigens presented on the cell surface induce an immune response. A study investigating the role of TP53 in human NSCLC cell lines also revealed that certain TP53 mutations lead to a direct, micro-RNA-mediated upregulation of PD-L1, suggesting a mechanism to dampen proinflammatory mutational effects [24]. In SGC, TP53 mutations occur at an overall frequency of 30% [25], with higher frequency in more aggressive carcinomas. Additionally, frequency of TP53 mutations correlated with the tumor mutational burden, which indicates a higher immunogenic potential [26]. In SGC as in several other carcinomas, TP53 mutations have been associated with a significantly shorter survival [27].

For the first time, the present study depicts the expression of LAG3 in a large cohort of SGC considering different well-defined subtypes in order to forge new paths for effective systemic therapy.

LAG3, CD8 and TP53 expressions per carcinoma type are displayed in Table 1. The relationship between the above-mentioned markers and clinical characteristics for the validation cohort are presented in Table 2. Exemplary multicolor IHC stainings are shown in Fig. 1. Survival data can be taken from Table 3, while the estimates of the 10-year EFS by the Kaplan–Meier model are shown in Figs. 2 and 3. Interdependencies for LAG3, CD8 and TP53 are demonstrated in supplementary Table 1.

Therapeutic options for metastatic and recurrent SGC are still sparse, even though several studies have shed light into their mutational landscape: Few recurrent actionable genomic alterations have been reported in certain subtypes, but most patients cannot be offered a targeted therapy nowadays. While other tumors have been extensively profiled for the expression of immunosuppressive molecules in the past years, only a few studies have depicted the TME of SGC. To our knowledge, we are the first to describe the importance of LAG3 in a bicentric approach with a test cohort and a large validation cohort of SGC with a significant number of cases across the different histotypes.

In our exploratory test cohort, 30.3% of all SGCs were infiltrated by LAG3-expressing TILs. Strikingly, this frequency was almost exactly confirmed by our validation cohort, which exhibited LAG3 expression on TILs in 28.2% of the cases. Even though the two cohorts revealed divergent frequencies for some tumor types, the prevalence of LAG3-expressing TILs in SDC, one of the most aggressive entities, was consistently high (50% and 66.7% for the test cohort and the validation cohort, respectively).

Our validation cohort can be separated in two prognostic groups: SDC and ANOS are often locally advanced tumors with lymph node metastasis and relatively poor prognosis. Tumors of other histologic groups which were included in the cohort tend to grow more indolently (e.g., MEC). While only approximately 20% of the latter were LAG3 positive, the former, aggressive tumor types were infiltrated by LAG3-positive TILs at high frequencies of 66.7% and 50%, respectively. This has direct translational relevance, as among others, patients with SGC are currently eligible for two clinical studies which evaluate the potential of LAG3 blockage either alone or in combination with nivolumab (NCT01968109, NCT02720068). For SDC and ANOS, new approaches for targeted therapies are warranted: While SDC is frequently HER2 (ERBB2) amplified and recent clinical studies involving trastuzumab yielded promising results, response duration is limited [31]. In contrast, ANOS lacks elsewhere-prevalent oncogenic mutations, leaving cytotoxic regimen as sole therapeutic option for recurrent or metastatic cases [25].

But SGC types with more variable aggressiveness, for instance AdCy, can also recur, metastasize or be irresectable. For these tumor groups, the therapeutic blockage of LAG3 could be considered as well.

Previous studies have demonstrated that SGCs express PD-L1, PD-L2 and CTLA at varying frequencies of up to 86% [12, 13, 15, 16]. Also, checkpoint inhibitor therapy leads to dramatic responses in some patients [18, 20]. Together with the mentioned results, our findings suggest that inhibition of immune-checkpoint molecules could be an option for systemic therapy.

IRs are expressed in the complex interplay between tumor cells and the immune system. Many studies contributed to the concept that chronic, tumor cell-controlled inflammation results in a favorable micro-environment for tumor progression [32]. Among many other parameters, infiltration by CD8-positive T lymphocytes and TP53 mutations was reported to play an important role in this context. Our data demonstrate that nonsense mutations (NT) of TP53 lead to a higher frequency of LAG3 positivity compared to WT TP53. This loss of TP53 protein function results in higher immunogenicity due to an accumulation of somatic mutations. Furthermore, loss of TP53 leads to an upregulation of another IR, PD-L1, through the lack of micro-RNAs miR-34a and miR-200 [24]. While cases with missense mutations (OT) were LAG3 positive in fewer cases than nonsense mutations (NT), their extent of LAG3-bearing lymphocytes was higher (36.8%, 16.7% and 9% for OT, NT and WT, respectively). Missense mutations have been demonstrated to possess gain-of-function (GOF) properties, possibly resulting in very different forms of immunomodulation [33, 34]. Additionally, as shown in HNSCC, neoantigens derived from mutated TP53 itself are detected by cytotoxic and helper T cells [35]. We hypothesize that the different immunomodulatory potential of missense mutations might lead to a more variable LAG3 expression than nonsense mutations. Whether certain GOF mutations or loss of TP53 could have direct, distinct impacts on LAG3 regulation should be further investigated.

An increasing CD8 expression has widely been correlated with antitumor immune response and tumor regression. Further, expressions of IRs and CD8 correlate in a variety of tumors, suggesting a means to dampen antitumor inflammatory responses [36]. In the present study, CD8-positive immune infiltration strongly correlated with LAG3 expression (p < 0.001), indicating that also in SGCs LAG3 expression might be a means to control an antitumor immune response. In conclusion, the results of our interdependency analysis show that LAG3 expression on TILs is upregulated in a setting of TP53 mutations with different LAG3 expression patterns for nonsense and missense mutations.

Our multicolor stainings revealed a predominant co-expression of LAG3 and CD8 in TILs. Several studies demonstrated LAG3 expression on effector CD8+ TILs in different cancer types [7, 37]. Apparently, LAG3+ exerts immunosuppressive functions on CD8+ TILs even in the absence of CD4+ helper T cells [38]. Furthermore, we observed LAG3 expression on both CD4+ FOXP3+ and CD4+ FOXP3− T cells. While the former have been termed immunosuppressive Tregs, LAG3 expression on both immunophenotypes has been linked to inhibition of antitumor activity [5, 39]. We demonstrated intratumoral LAG3 expression in both the T helper and the effector T cell compartment. Studies concerning other tumor entities are in line with this finding and have revealed evidence for a functional impact of this expressional pattern. Thus, our data indicate that LAG3 plays an important role in the immunological TME of SGC.

In the univariate analysis, the expression of LAG3 did not significantly alter EFS for the overall cohort. Contrarily in the entity-wise univariate analysis, LAG3 positivity correlated with a decreased EFS in AdCy, while no prognostic effect was detected for the remaining histologic groups. This prognostic relevance is in line with previous studies investigating LAG3 in other tumor types [8, 40], even though studies with breast cancer came to divergent results [41, 42]. Using a multivariate cox analysis, we tested whether LAG3 expression might be a predictor of decreased EFS independent of the covariates age category, N- and T-stage, therapy, consumption of alcohol and nicotine. Indeed, we found a trend for shorter EFS in the LAG3-positive subgroup of AdCy. For the remaining variables, CD8 expression and TP53 mutation, we previously demonstrated a significant interdependency with LAG3 and consecutively excluded these variables from multivariate analysis. Hence, we cannot rule out that the prognostic effect of LAG3 expression could be partially attributed to these two cofactors.

Even though the impact of LAG3 expression on EFS was confined to the AdCy subgroup, we demonstrated that in the overall cohort, LAG3 expression significantly correlated with higher N-stage and UICC classification. This suggests that LAG3 might have a significant prognostic relevance also for other histotypes in a larger cohort.

Prognostic implications of different mutational patterns of TP53 expression are sparsely documented. While TP53 mutations are widely regarded as predictors for a poor outcome, in most studies missense and nonsense mutations do not differ significantly regarding outcome [43]. We found only one study that reported a differential prognosis for nonsense compared to missense mutations: In this publication, nonsense mutations in SDC were associated with shorter progression-free survival than missense mutations [44]. Similarly, in our overall cohort, we demonstrated that TP53 nonsense mutations are associated with a shorter EFS in univariate analyses (Kaplan–Meier model, univariate cox regression). Additionally, the multivariate cox regression analysis revealed a trend (p = 0.065) toward a shorter EFS for these mutations. Analogous to the above-mentioned procedure, here, we excluded the cofactor LAG3 for which we previously demonstrated a significant interdependency with TP53 mutations. Interestingly, we did not observe a comparable effect on the EFS for missense mutations, implicating profound differences of these mutation types for cancer aggressiveness. Comparable to our findings concerning LAG3 expression, TP53 mutations significantly correlated with higher T, N and UICC stage. This indicates that also TP53 mutations might have a more pronounced prognostic relevance on EFS in a larger dataset.

Due to its retrospective and descriptive nature, our study cannot yield evidence for effectiveness of LAG3 inhibition in SGC. To investigate its therapeutic potential, also with respect to tumor mutational burden as a possible predictive marker, clinical studies are warranted. We did not examine the expression of other immunological markers such as PD-L1, CTLA-4 and TIM-3. Whether the co-expression of LAG3 with these molecules could have a more pronounced prognostic impact should be addressed by future studies. This could be of particular interest as synergistic LAG3/PD-L1 co-expression has been associated with T cell anergy [7]. Moreover, multiple trials evaluate the combinatory inhibition of LAG3 and other IRs [45].