Factors that increase class I MHC expression may contribute to the development of immune checkpoint inhibitor-induced diabetes

Abstract

Immune checkpoint inhibitors (ICIs) have improved survival of patients with cancer, yet they pose risks of immune-related adverse events (irAEs). ICI-induced insulin-dependent diabetes mellitus (ICI-DM) is a life-threatening and life-altering irAE. Previously, we reported a high incidence of a germline missense variant in NLRC5, a key class I transcription activator, among patients with ICI-DM compared with similarly treated patients who did not develop ICI-DM. Our purpose was to validate this finding in additional ICI-treated patients and study effects of the NLRC5 variant on expression of class I major histocompatibility complex (MHC) antigen presentation genes.

We assessed the prevalence of the C>T missense variant at chr16:57 025 515 (NLRC5Pro191Leu) in germline DNA from an additional 33 patients with ICI-DM and in patients with ICI-induced colitis (n=15), ICI-induced hypothyroidism (n=19) and ICI-induced hypophysitis (n=17). The 1,000 Genomes Project was used for comparison. We assessed peripheral blood mononuclear cells from 16 individuals with or without the NLRC5 variant, studying expression of NLRC5 and select downstream target genes, before and after stimulation with interferon-γ.

We validated the higher prevalence of NLRC5Pro191Leu in a non-overlapping cohort of patients with ICI-DM compared with the general population (51.5% vs 12.8%, p<0.0001). The prevalence of NLRC5Pro191Leu in ICI-induced colitis or thyroiditis patients did not significantly differ from the general population, while the prevalence in ICI-induced hypophysitis was somewhat higher (21.6%, p=0.048). We found greater increases in messenger RNA expression of NLRC5 (p=0.007), TAP1 (p=0.0002), B2M (p=0.0005), HLA-G (p=0.04), PSMB8 (p=0.03) and PSMB9 (p=0.01) in NLRC5Pro191Leu cells stimulated with interferon-γ compared with NLRC5^WT cells. A similar trend was observed for HLA-A (p=0.09).

We confirm the significantly higher prevalence of the NLRC5Pro191Leu variant in patients with ICI-DM relative to the general population. This abundance appears to be unique to patients who develop ICI-DM or hypophysitis on ICIs, underscoring its potential involvement in the pathogenesis of these endocrinopathies. The effects of this NLRC5 variant on class I MHC regulators suggest a mechanistic connection between the variant and development of ICI-DM. Further work is warranted to determine whether class I MHC molecules can be modulated in patients with the NLRC5Pro191Leu variant requiring ICIs.

Introduction

Immune checkpoint inhibitors (ICIs) play a pivotal role in treating various malignancies as they have significantly improved patient survival. However, T-cell activation in patients treated with ICIs is not limited to the tumor microenvironment, and immune-related adverse events (irAE) have been observed in almost every organ.^{1 2} While most irAEs are reversible with immune suppression and possibly with cessation of ICIs, ICI-induced endocrine irAEs are typically irreversible. ICI-induced hypophysitis can be deadly if not diagnosed and treated appropriately, and ICI-induced pancreatic β-cell destruction causing insulin-dependent diabetes mellitus (ICI-DM) is similarly irreversible and resultant diabetic ketoacidosis (DKA) and hyperglycemia can be life-threatening. Risk factors for and mechanisms of these severe, irreversible irAEs are a subject of intense research.^{3 4}

ICI-DM is caused by total and rapid β-cell destruction.⁵ The resultant brittle diabetes occurs in 0.2–2% of patients treated with ICIs and is more frequent in regimens containing the combination of anti-cytotoxic T-lymphocyte associated protein 4 (CTLA-4) and anti-programmed cell death protein-1 (PD-1).6,9 Although the precise mechanisms by which ICI-DM occurs remain unclear, ICI-mediated T-cell activation triggers autoimmune β-cell destruction.^{10 11} To the best of our knowledge, β-cell destruction cannot be halted or reversed by corticosteroids, although this has not been extensively used due to the risk of worsening hyperglycemia.⁹ As the use of ICIs expands, including their utilization in adjuvant therapies, the incidence of this lifelong complication is rising, significantly impacting clinical outcomes and quality of life.¹²

We previously performed whole exome sequencing of germline DNA from 13 patients with ICI-DM, and found that 9 (69.2%) harbored a missense variant (C>T) at chr16:57 025 515 in the NLR family CARD domain containing 5 (NLRC5) gene resulting in a proline to leucine substitution (NLRC5Pro191Leu rs74439742 in the dbSNP database).¹³ We compared these 13 patients to age-matched and gender-matched patients with the same tumor and treatment type who did not develop ICI-DM and did not find the NLRC5Pro191Leu variant in any.¹³ In 694 individuals sequenced in a general US population in the 1,000 Genomes Project,¹⁴ we found a substantially lower prevalence of 12.8% (p=5.98×10⁻⁶).¹³ We assessed the prevalence of the NLRC5Pro191Leu variant in patients with type 1 diabetes not induced by ICIs and the variant was not associated with other types of type 1 diabetes.¹³

The NLRC5 gene, located on the 16q13 locus, is expressed in various tissues, with particularly high expression levels in immune cells and mucosal epithelia, but not in the pancreas. Analogous to CIITA (class II major histocompatibility complex (MHC) transactivator gene), which induces MHC class II gene expression, NLRC5 promotes the expression of MHC-I genes, thus called MHC-I transactivator (CITA).^{15 16} NLRC5 induces expression of genes coding for B2M, PSMB9, and TAP1, all of which are involved in antigen processing and presentation to CD8+T cells.^{17 18} Proteasome subunit beta types 8 (PSMB8) and human leukocyte antigen-A (HLA-A) are also co-expressed with NLRC5.¹⁹ Due to its role as a transactivator of MHC-I and its influence on expression of key components involved in antigen presentation, NLRC5 serves as a central regulator of immune-mediated responses.²⁰

In our previous studies, we showed that the NLRC5^Pro191Leu missense variant was recurrent in all patients studied.¹³ In tumors sequenced in our previous studies, no additional missense variants were found, but some patients had non-synonymous tumor mutations, which by nature would not induce immunity. Variations in the NLRC5 gene and its phenotype may be linked to a wide range of immune responses. NLRC5 is a key target for antitumor immunity and immune evasion. Notably, NLRC5 upregulation was significantly linked to the activation of CD8+cytotoxic T cells, response to treatment and improved patient survival across multiple cancer types.¹⁶ However, the effects of the NLRC5Pro191Leu variant on downstream molecules are unknown.

Our purpose was to verify our initial observation of enrichment of the NLRC5^Pro191Leu variant in a larger cohort of patients with ICI-DM and to explore its functional downstream effects.

Methods

Patient selection

With approval of Institutional Review Boards of Yale University and Massachusetts General Hospital, and after obtaining written consent from patients, we collected tissue or peripheral blood samples (PBMCs) from 33 patients diagnosed with ICI-DM (22 from Yale University and 11 from Massachusetts General Hospital). Five patients without available PBMCs for research had sequencing done from a pelvic metastasis, a uterine adenocarcinoma, a rectal polyp, a small cell lung carcinoma sample, and a pleural metastasis from lung adenocarcinoma. Tumor samples from 15 patients with ICI-induced colitis and blood samples from 19 with ICI-induced thyroiditis and 74 diagnosed with ICI-induced hypophysitis were used for comparison.

DNA extraction and sequencing

For tissue samples, H&E slices were reviewed and 1 mm core samples were collected for DNA extractions. DNA extraction from blood or tissue was conducted using a Qiagen DNEasy Blood & Tissue Kit. We performed PCR followed by Sanger sequencing to identify NLRC5^Pro191Leu variant, employing primers designed with primer-BLAST and used to amplify the region of interest: Forward:5'-TTGGCCAAGAAGTACCTGCAGC-3'; Reverse:5'-TCGTGATCAAGTTGAGCTGGCG-3' (online supplemental figure S1). For patients with colitis, existing whole exome sequencing data obtained for other purposes was used. Whole exome sequencing was performed by the Yale Center for Genomic Analysis. Data were processed by the genome analysis toolkit variant calling and documented with ANNOtate VARiation (ANNOVAR) software .

Peripheral blood mononuclear cell stimulation and gene expression analysis

PBMCs from individuals with NLRC5^Pro191Leu and NLRC5^WT were plated in 96-well dishes (0.2×10⁶ cells per well) and treated with 10 ng/mL interferon (IFN)-γ (R&D). Cells were harvested after 24 hours. RNA was isolated using the Qiagen RNeasy Plus Micro Kit (Hilden, Germany), converted to complementary DNA (cDNA) (High-Capacity cDNA Reverse Transcription Kit, Applied Biosystems, Thermo Fisher) and quantitative reverse transcription polymerase chain reaction (qRT-PCR) performed using iTaq Universal SYBR Green Supermix (Bio-Rad). The following primers were used: NLRC5 Forward: GCTCGGCAACAAGAACCTGT, Reverse: GGTCCAAGGTCTCGTTCCT; TAP1 Forward: CGCCTCACTGACTGGATTCTA, Reverse: TCTGTTGGAAAAACTCCGTCTC; B2M Forward: TATCCAGCGTACTCCAAAGA, Reverse: GACAAGTCTGAATGCTCCAC; HLA-A Forward: AAAAGGAGGGAGTTACACTCAGG, Reverse: GCTGTGAGGGACACATCAGAG; HLA-G Forward: GAGGAGACACGGAACACCAAG, Reverse: GTCGCAGCCAATCATCCACT; PSMB8 Forward: CACGCTCGCCTTCAAGTTC, Reverse: AGGCACTAATGTAGGACCCAG; PSMB9 Forward: GGTTCTGATTCCCGAGTGTCT, Reverse: CAGCCAAAACAAGTGGAGGTT. The ACTB housekeeping gene was used for normalization and gene transcription presented as ΔCt=ΔCt=CtACTB−Ct Target Gene. Gene expression was calculated as 2^ΔCt. Data were analyzed with GraphPad Prism V.10 (La Jolla, California, USA) and presented as mean±SEM.

Data analysis

Demographic comparisons were conducted using χ² tests for categorical variables and t-tests for continuous variables. The Kaplan-Meier method was used to compare survival and cumulative risk. Differences in downstream molecular changes following IFN-γ stimulation between NLRC5Pro191Leu and NLRC5^WT were analyzed using the Mann-Whitney test. A p value of <0.05 was considered statistically significant. Data analysis was performed using R software (V.4.4.1).

Results

Patient with ICI-DM demographics and tumor characteristics

To validate our previously described findings in 13 patients with ICI-DM, an additional 33 patients were studied, 22 from Yale New Haven Hospital and 11 from Massachusetts General Hospital. Patient characteristics are provided in table 1. The mean age at cancer diagnosis was 71.3 years±10.3, 16 (48.5%) were men, and the majority were Caucasian. Due to early use of ICIs in certain malignancies, most patients (63.6%) had melanoma. The remaining had renal cell carcinoma, adenocarcinoma of the lung, small cell lung cancer, gastroesophageal adenocarcinoma, pancreatic cancer, endometrial cancer or breast cancer. Combination anti-PD-1 and anti-CTLA-4 therapy was administered to 51.5%, and the remainder received anti-PD-1/programmed death-ligand 1 (PD-L1) monotherapy. At ICI-DM diagnosis, 22 (66.7%) presented with DKA and the remainder with rapid onset hyperglycemia. Tumor response to immunotherapy was evenly distributed, with 45.6% (n=21) showing a favorable response. Demographics of the initial cohort of 13 patients were previously described.¹³

Table 1. Demographic and clinical characteristics of the validation ICI-DM cohort (n=33).

Age at cancer diagnosis (years)
Mean	71.3
SD±*	±10.3
Gender
Male	16 (48.5%)
Female	17 (51.5%)
Race
White or Caucasian	30 (90.1%)
Unknown	3 (9.1%)
Ethnicity
Non-Hispanic	29 (87.9%)
Hispanic or Latino	3 (9.1%)
Unknown	1 (3.0%)
Type of cancer
Melanoma	21 (63.6%)
Renal cell carcinoma	4 (12.1%)
Lung adenocarcinoma	2 (6.1%)
Other	6 (18.2%)
Type of ICI†
Anti PD-1 and anti-CTLA-4	17 (51.5%)
Anti PD-1/PD-L1	16 (48.5%)
Other ICI†-related toxicities (Grade ≥3)
Yes	25 (75.8%)
No	8 (24.2%)
Time from treatment initiation to diabetes (months)
Median	6.2 (95% CI: 6.7 to 15.4)
DKA‡ at diabetes diagnosis
Yes	22 (66.7%)
No	11 (33.3%)
Response to immune checkpoint inhibitors
Complete or partial response	14 (42.4%)
Stable disease	5 (15.1%)
Progressed disease	8 (24.2%)
Not applicable or mixed response	6 (18.2%)

^*SD.

^†ICI: immune checkpoint inhibitors.

^‡ DKA: diabetes ketoacidosis.

CTLA-4, cytotoxic T-lymphocyte associated protein 4; ICI-DM, immune checkpoint inhibitor-induced insulin-dependent diabetes mellitus; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1.

NLRC5 sequencing in the validation ICI-DM cohort

The NLRC5Pro191Leu variant was found in 17 (51.5%) of the 33 validation cohort patients, two homozygous. While the incidence is lower than in the previous cohort of 13 patients, the difference was not statistically significant (χ² p=0.28), figure 1A. The incidence in the validation cohort also remained significantly higher than the general population, where the NLRC5Pro191Leu variant is seen in 12.8% (p<0.0001), figure 1B. The likelihood of developing ICI-DM in individuals with NLRC5Pro191Leu who develop ICI-treated cancer is 7.1-fold (95% CI: 3.3 to 15.8) that of NLRC5^WT individuals, χ² p=34.8.

Figure 1. (A) Prevalence of NLRC5Pro191Leu variant in patients with ICI-DM: previously published discovery cohort (n=13) versus validation cohort (n=33). (B) Prevalence of NLRC5Pro191Leu in patients with ICI-DM: validation cohort (n=33) versus general population (n=694). (C) Prevalence of NLRC5Pro191Leu in patients with various immune-related adverse events: ICI-DM (n=33), colitis (n=53), hypothyroidism (n=19), hypophysitis (n=79) versus general population (n=694). ICI-DM, immune checkpoint inhibitors-induced insulin-dependent diabetes mellitus; NLRC5, NLR family CARD domain containing 5.

Timing of ICI-DM onset and overall survival

We next determined whether patients with ICI-DM with NLRC5Pro191Leu developed earlier onset ICI-DM than NLRC5^WT and found no significant difference; 6.2 months (95% CI: 1.5 to 13.4) versus 6.4 (95% CI: 1.9 to 32.6), respectively, (p=0.21), figure 2A. Patients in this cohort had various tumor types and were not uniformly treated. We therefore focused on melanoma only for survival analyses, figure 2B. The median overall survival (OS) for all patients with melanoma patients was 85.0 months (95% CI: 48.2 to Not Reached [NR]). NLRC5Pro191Leu patients had a median OS of 99.2 months (95% CI: 48.2 to NR), compared with 54.0 months (95% CI: 34.0 to NR) for those with NLRC5^WT (p=0.098).

Figure 2. (A) Time from immune checkpoint initiation to immune checkpoint inhibitors-induced insulin-dependent diabetes mellitus diagnosis in NLRC5Pro191Leu (n=17) versus NLRC5WT (n=16) patients. (B) Kaplan-Meier 5 years survival analysis of patients with melanoma with NLRC5Pro191Leu (n=10) versus NLRC5WT (n=11) patients. NLRC5, NLR family CARD domain containing 5.

NLRC5 sequencing in DNA from patients with other irAEs

We then determined whether the NLRC5Pro191Leu variant is also more prevalent in patients with other select irAEs. Among patients with colitis and thyroiditis/hypothyroidism, the prevalences of the NLRC5Pro191Leu were not significantly higher than in the general population (26.6%, p=0.12 and 26.3%, p=0.09, respectively). The incidence of NLRC5Pro191Leu among patients who developed ICI-induced hypophysitis was 21.6%, higher than the general population, p=0.048, figure 1C.

Effects of the NLRC5^Pro191Leu variant on expression of class I MHC molecules

Given the association between the NLRC5Pro191Leu variant and ICI-DM, we sought to determine whether the variant influences class I MHC expression. We cultured PBMCs from eight healthy NLRC5Pro191Leu individuals and eight NLRC5wt. Each experiment was performed in triplicate, assessing the expression of NLRC5 and six downstream genes regulated by NLRC5 (online supplemental figure S1), before and after IFN-γ exposure. The fold change in gene expression was calculated and averaged across triplicate experiments to compare the two groups. Cells harboring the NLRC5Pro191Leu variant exhibited significantly greater increases in messenger RNA expression of NLRC5 (p=0.007), TAP1 (p=0.0002), B2M (p=0.0005), HLA-G (p=0.04), PSMB8 (p=0.03) and PSMB9 (p=0.01), figure 3A-F. TAP1 exhibited the most pronounced upregulation, with approximately a threefold increase in expression after IFN-γ stimulation in NLRC5Pro191Leu cells relative to NLRC5^WT. Although HLA-A expression was also somewhat higher in NLRC5Pro191Leu cells, the difference did not reach statistical significance (p=0.09). Pearson correlation analyses revealed that NLRC5 expression positively correlates with all other downstream mediators studied, with stronger correlations observed in PBMCs carrying the NLRC5Pro191Leu variant compared with those with NLRC5^WT (figure 3H, online supplemental figure S2).

Figure 3. (A–G) Fold change in messenger RNA expression of (A) NLR family CARD domain containing 5 gene, NLRC5, (B) Transporter Associated with Antigen Processing 1 gene, TAP1, (C) Beta-2 microglobulin gene, B2M, (D) Human Leukocyte Antigen-A, HLA-A, (E) Human Leukocyte Antigen-G, HLA-G, (F) Proteasome Subunit Beta Type-8, PSMB8, and (G) Proteasome Subunit Beta Type-9, PSMB9 in patients with NLRC5^Pro191Leu and NLRC5^wt following interferon-γ stimulation. (H–I) Correlation between expression of NLRC5^WT (G) or NLRC5^Pro191Leu (H) and downstream genes.

Discussion

We previously identified a missense variation in the NLRC5 gene, which was significantly more prevalent in 13 patients who developed diabetes when treated with ICIs, compared with the general population.¹³ Here we validated the enrichment of the chr16:57025515 C>T variant in a larger, non-overlapping cohort of 33 patients, demonstrating a variant prevalence of 51.5% in patients with ICI-DM, significantly higher than the general population (p=2.90×10⁻⁷). Importantly, individuals with the NLRC5Pro191Leu variant are 7.1 times more likely to develop diabetes when treated with ICIs than individuals without.

The role of NLRC5 in class I MHC regulation and antigen presentation suggests that the NLRC5Pro191Leu variant might mechanistically be associated with the development of irAEs in the setting of intense inflammation, such as ICIs. Here we demonstrated that PBMCs harboring the NLRC5Pro191Leu have higher transcription of multiple class I MHC regulators on exposure to IFN-γ. These findings underscore the potential clinical significance of NLRC5Pro191Leu in immune-mediated toxicity, particularly in the context of intense inflammation induced by ICIs. This variant may enhance the recognition of self-antigens by CD8⁺ T cells, potentially triggering an autoimmune response that leads to β-cell destruction.

Higher expression of class I MHC molecules might therefore lead to increased incidence of other irAEs. We found a weaker but statistically significant link between the NLRC5Pro191Leu variant and ICI-induced hypophysitis, suggesting that a common mechanism of endocrine cell destruction might be at play with these two endocrine irAEs in a subset of patients. While the NLRC5Pro191Leu did not appear to be enriched among patients with ICI-induced thyroiditis or colitis, the numbers of patients with these two irAEs were small. Overall, however, the substantially greater enrichment of this variant among patients with ICI-DM suggests that the mechanism driving pancreatic β-cell destruction leading to ICI-DM might be different than that of the other irAEs assessed here.

Previous studies have demonstrated an association between NLRC5 upregulation and improved survival across various cancer types.^{19 21} Given that NLRC5 expression levels were elevated in the presence of the NLRC5^Pro191Leu variant, we hypothesized that it might be associated with improved survival in patients with cancer treated with ICIs. While a similar trend was observed, we did not find a statistically significant improvement in survival among patients with melanoma with the investigated variant, possibly attributable to the small sample size and admixture treatment regimens.

In the NLRC5 protein, residue 191 is positioned between the N-terminal and NACHT domains online supplemental figure S3, which play key roles in interacting with junction proteins, activating downstream signaling, and facilitating nuclear-cytoplasmic shuttling.²² The transactivation of MHC I by NLRC5 depends on its structural integrity, and the proline-to-leucine substitution at this position may introduce structural alterations that affect protein function. However, the specific roles of NLRC5 domains remain poorly characterized, highlighting the need for further research.

Although ICIs result in long-term benefit in many patients with cancer, treatments that harness alternative mechanisms are becoming available. Identifying a genetic predisposition to ICI-DM can affect treatment selection, especially in settings in which ICIs are less effective. Moreover, some studies have shown higher rates of ICI-DM when anti-PD-1 is given together with anti-CTLA-4.²³ Individuals with the NLRC5Pro191Leu variant may consider a regimen that does not include anti-CTLA-4. With increasing use of anti-PD-1/L1 in the adjuvant setting for various malignancies, patients at risk for ICI-DM, such as those with NLRC5Pro191Leu, may consider avoiding anti-PD-1 if they can be cured by surgery alone. Finally, as select immune-suppressive drugs are being incorporated into treatment regimens for patients with cancer at risk for irAEs, patients at greater risk for ICI-DM can be monitored closely for early signs of ICI-DM to avoid total β-cell destruction, including the use of home glucometers. In addition to guiding treatment selection, the observed trend toward improved survival in patients with melanoma harboring the variant suggests it may also have prognostic significance. This requires further validation.

This study has a number of limitations. Although a larger validation cohort was interrogated, the sample size remains small due to the low incidence of ICI-DM and limited to two academic medical centers. Larger studies are essential to validate our findings in more diverse tumor types and immunotherapy regimens. Further exploration of the molecular mechanisms of NLRC5Pro191Leu-associated ICI-DM is a necessary step towards clinical application of these findings.

In conclusion, we validated our initial finding linking the chr16:57025515 C>T NLRC5 variant to susceptibility to developing ICI-DM in a separate and somewhat larger cohort of patients from two institutions. The high prevalence (>50%) of the NLRC5Pro191Leu variant in patients with ICI-DM, together with evidence of altered downstream transcription, supports its role as a genetic predisposing factor, potentially through mechanisms involving altered class I MHC regulation and antigen presentation, and these changes appear to be most important with IFN-γ stimulation. This finding has important clinical implications given the increasing use of immune checkpoint inhibitors in patients with cancer.

Data availability statement

Data are available upon reasonable request.