Early reduction in PD-L1 expression predicts faster treatment response in human cutaneous leishmaniasis

Abstract

Cutaneous leishmaniasis (CL) is caused by Leishmania donovani in Sri Lanka. Pentavalent antimonials (e.g., sodium stibogluconate [SSG]) remain first-line drugs for CL with no new effective treatments emerging. We studied whole blood and lesion transcriptomes from Sri Lankan patients with CL at presentation and during SSG treatment. From lesions but not whole blood, we identified differential expression of immune-related genes, including immune checkpoint molecules, after onset of treatment. Using spatial profiling and RNA-FISH, we confirmed reduced expression of programmed death-ligand 1 (PD-L1) and indoleamine 2,3-dioxygenase 1 (IDO1) proteins on treatment in lesions of a second validation cohort and further demonstrated significantly higher expression of these checkpoint molecules on parasite-infected compared with noninfected lesional CD68+ monocytes and macrophages. Crucially, early reduction in PD-L1 but not IDO1 expression was predictive of rate of clinical cure (HR = 4.88) and occurred in parallel with reduction in parasite load. Our data support a model whereby the initial anti-leishmanial activity of antimonial drugs alleviates checkpoint inhibition on T cells, facilitating immune-drug synergism and clinical cure. Our findings demonstrate that PD-L1 expression can be used as a predictor of rapidity of clinical response to SSG treatment in Sri Lanka and support further evaluation of PD-L1 as a host-directed therapeutic in leishmaniasis.

Keywords: Immunology; Immunotherapy; Infectious disease; Molecular pathology; Parasitology.

Figure 1. Differential expression and network analysis of genes regulated by drug treatment in lesions of Sri Lankan patients with CL.

Immune-targeted tissue transcriptomics were conducted on tissue sections from test cohort patients comparing transcriptomes at presentation and on treatment. (A) Principal component analysis was performed to show differences between the pre- and on-treatment transcriptome of each patient based on 770 genes from the nCounter PanCancer Immunology Panel (n = 6). (B) DEGs comparing pretreatment biopsies with biopsies taken after 2 weeks on treatment (SSG). Cut off (red line) drawn at equivalent of adjusted P = 0.01 and log (fold change) of 1. (C) Top 30 genes that changed in expression on SSG treatment. (D) STRING protein-protein interaction network (https://string-db.org; ref. 22) analysis of genes listed in Supplemental Table 3 downregulated on SSG treatment. Pathways represent GO: 0072676, lymphocyte migration (red spheres), and GO: 0002684, positive regulation of immune system process (blue spheres). Top 20 genes are shown (log₂ fold change ≥ 1.15) for clarity.

Figure 2. Digital spatial profiling of CL lesions.

Digital spatial profiling was performed on tissue sections from test cohort individuals comparing ROIs from pre and on-treatment biopsies. (A–F) ROIs on CD3⁺ and/or CD68⁺ rich areas from pre- and on-treatment biopsies from patients P4, P6, and P7 (CD68, green; CD3, red; Syto13, blue). Original magnification: ×20; scale bar: 500 μm. (G) t-SNE plot based on 20 PCA loadings colored on patient ID. (H) Differential protein expression analysis comparing pretreatment to on-treatment ROIs. Red lines indicate adjusted P value cut off of 1% (Mann-Whitney test with FDR correction based on Benjamini, Krieger, and Yekutieli 2-stage set-up method) and log₂ fold change = 0.5 (n = 33 ROIs). (I) IDO1, PD-1, and PD-L1 expression in pre- and on-treatment ROIs. Mann Whitney rank test (n = 33 ROIs). (J) Immunohistochemistry (IHC) was performed on sections from patients in the validation cohort (n = 23) before and during treatment, and quantitated using StrataQuest (see Methods). Comparisons were by Wilcoxon matched-pairs signed rank test. Dotted lines show upper and lower quantile in I and J, median by solid line.

Figure 3. Immunofluorescence analyses of PD-L1 in infected and uninfected cells.

Dual IHC-FISH using an Amastin probe was performed on pretreatment sections of patients enrolled in the validation cohort. (A) A 400× confocal image showing infection of PD-L1⁺CD68⁺ (arrows) and PD-L1^–CD68⁺ (arrowhead) cells. Scale bar: 50 pixels. (B) Relationship between PD-L1 expression and parasite burden (Amastin dot count). Scattergram from a representative patient (P24 at presentation) showing Amastin⁺ low (cyan), medium (red), and high (green) PD-L1–expressing cells with respect to parasite abundance. (C) Fluorescence intensity distributions of infected and uninfected PD-L1 cells. (D) Mean fluorescence intensity of PD-L1 expression on Amastin^– cells compared with Amastin⁺ cells from representative patient P24. The upper and lower whisker represents highest and lowest value that is within 1.5 times the interquartile range. n = 9159 parasite positive cells and n = 41520 parasite negative cells. Significance score was generated using Wilcoxon signed rank test. (E) PD-L1 expression on Amastin⁺PD-L1⁺ cells versus Amastin^–PD-L1⁺ cells (n = 7 patients). Significance score was generated using Student’s 2-tailed paired t test after testing for normality using Shapiro-Wilk and Kolkogorov-Smirnov tests.

Figure 4. Clinical correlates of PD-L1 reduction on treatment in patients with CL.

(A) Patients (validation cohort; n = 23) were stratified based on high (> geomean value; n = 11) and low (< geomean value; n = 12) pre-/on-treatment expression ratio. (B) Kaplan-Meier curve based on pre-/on-treatment ratio of PD-L1 expression (high versus low). (C) Patients stratified based on on-treatment expression of PD-L1 (> geomean value; n = 11 versus < geomean value; n = 12). (D) Multivariate Cox proportional hazards model plotted as a forest plot. P values for each covariate represent Wald statistic value, and overall statistical significance is also indicated. (E) Patients stratified by LITS1 PCR status (n = 9 PCR⁺ versus n = 14 PCR^– or PCR^+/– [equivocal]) on treatment. (F) PD-L1 expression in LITS PCR⁺ versus PCR^– individuals on treatment. Dotted lines show upper and lower quantile, solid line shows median. P value generated using 2-tailed Mann-Whitney test. Vertical line drawn in B, C, and E on the x axis shows time when on-treatment biopsies were collected. Curves in B, C, and E were compared using log-rank (Mantel-Cox) test. Blue and red shaded areas show 95% CI of the 2 groups.