Immune checkpoint inhibitor-associated celiac disease

Abstract

Background

Rare cases of immune checkpoint inhibitor (ICI)-associated celiac disease (ICI-CeD) have been reported, suggesting that disruption of tolerance mechanisms by ICIs can unmask celiac disease (CeD). This study aims to characterize the clinicopathological and immunophenotypic features of ICI-CeD in comparison to ICI-associated duodenitis (ICI-Duo) and usual CeD.

Methods

A medical and pathological records search between 2015 and 2019 identified eight cases of ICI-CeD, confirmed by tTG-IgA. Nine cases of ICI-Duo, 28 cases of moderate CeD, as well as 5 normal controls were used as comparison groups. Clinical information was collected from the electronic medical records. Immunohistochemistry for CD3, CD8, T-cell receptor gamma/delta (γδ), programmed death ligand 1 (PD-L1), and programmed death 1 (PD-1) were performed, with quantification of intraepithelial lymphocyte (IEL) subsets in three well-oriented villi. CD68, PD-L1, and PD-1 were assessed as a percentage of lamina propria surface area infiltrated by positive cells. Statistical significance was calculated by the Student’s t-test and Fisher’s exact test.

Results

The eight patients with ICI-CeD (F:M=1:3) and nine patients with ICI-Duo (F:M=5:4) presented similarly with diarrhea (13/17) and abdominal pain (11/17) after a median of 1.6 months on ICI therapy. In patients with ICI-CeD, tTG-IgA ranged from 104 to >300 IU/mL. Histological findings in ICI-CeD and ICI-Duo were similar and included expansion of the lamina propria, active neutrophilic duodenitis, variably increased IELs, and villous blunting. Immunohistochemistry showed that the average number of IELs per 100 enterocytes is comparable between ICI-CeD and ICI-Duo, with increased CD3 ⁺ CD8 ⁺ T cells compared with normal duodenum but decreased γδ T cells compared with CeD. Average PD-L1 percentage was 9% in ICI-CeD and 18% in ICI-Duo, in comparison to <1% in CeD and normal duodenum; average PD-1 percentage was very low to absent in all cases (<3%). On follow-up, five patients with ICI-CeD improved on a gluten-free diet (GFD) as the sole therapeutic intervention (with down-trending tTG-IgA) while the other three required immunosuppression. All patients who developed ICI-Duo received immunosuppression with variable improvement in symptoms.

Conclusions

ICI-CeD resembles ICI-Duo clinically and histologically but shares the serological features and response to gluten withdrawal with classic CeD. Immunophenotyping of IELs in ICI-CeD and ICI-Duo also shows similar CD3, CD8, γδ T cell subsets, and PD-L1 populations, all of which differed quantitatively from usual CeD. We conclude that ICI-CeD is biologically similar to ICI-Duo and is likely a variant of ICI-Duo, but treatment strategies differ, with ICI-CeD often improving with GFD alone, whereas ICI-Duo requires systemic immunosuppression.