Implications of Tuberculosis Reactivation after Immune Checkpoint Inhibition

To the Editor:

Treatment of malignant disease with immune checkpoint inhibitors is emerging as a transformative approach. However, tuberculosis (TB) reactivation associated with these agents is being increasingly reported (Table 1). We describe a further case of TB associated with anti–PD-1 (programmed cell death-1) immunotherapy and perform immunohistochemical analysis of lung biopsies from patients with standard or anti–PD-1–associated TB. We discuss the potential underlying mechanisms and implications for clinical practice and research.

Table 1.

Summary of Accumulating Evidence of Tuberculosis Reactivation with Anti–PD-1 Immune Checkpoint Inhibition

First Author	Year of Publication	Malignancy	Inhibitor	Site of TB	DOI
Fujita K	2016	Non–small cell lung cancer	Nivolumab	Pulmonary	10.1016/j.jtho.2016.07.006
Lee JJ	2016	Hodgkin's lymphoma	Nivolumab	Pulmonary	10.3109/0284186X.2015.1125017
Chu YC	2017	Non–small cell lung cancer	Nivolumab	Pericardial	10.1016/j.jtho.2017.03.012
Picchi H	2018	Melanoma	Pembrolizumab	Pleural	10.1016/j.cmi.2017.12.003
		Non–small cell lung cancer	Nivolumab	Spinal	10.1016/j.cmi.2017.12.003
Jensen KH	2018	Non–small cell lung cancer	Nivolumab	Pulmonary	10.1080/0284186X.2018.1433877
Elkington PT	2018	Melanoma	Pembrolizumab	Pulmonary and hepatic	Present work

Definition of abbreviations: PD-1 = programmed cell death-1; TB = tuberculosis.

TB Associated with Pembrolizumab, an Antibody to PD-1

A 62-year-old woman was diagnosed with ocular melanoma, which was excised. Three years later, metastatic disease developed and immune checkpoint inhibition therapy was commenced, initially with ipilimumab, an anti–CTLA-4 (cytotoxic T-lymphocyte–associated antigen 4) antibody, and then with pembrolizumab, a humanized monoclonal antibody against PD-1. The disease was stable for 2 years, but then blood liver biochemical markers became abnormal and a lung lesion was noted on computed tomographic scanning. Liver function abnormalities persisted despite immunosuppression, and so a liver biopsy was performed, which showed a single granuloma. Biopsy of the cavitating apical lung lesion showed necrotizing granulomatous inflammation, and bronchial washings cultured Mycobacterium tuberculosis (Mtb). Antituberculosis treatment was initiated, which led to clinical improvement, normalization of liver function tests, and regression of the lung lesion. Therefore, the unifying diagnosis was disseminated TB associated with immune checkpoint inhibition. However, this clinical occurrence runs counter to the current disease paradigm, which proposes that active TB results from a deficient host immune response (1).

Therefore, we performed immunohistochemical analysis of TB lung lesions in the context of a normal immune response (six cases) and the lung biopsy of this case. Immunostaining was performed for PD-L1 (programmed cell death ligand-1), CD8, and PD-1. In normal TB granulomas, PD-L1 is highly expressed, while PD-1 colocalizes with CD8, demonstrating that immune checkpoint ligands and receptors are coexpressed (Figure 1A). In the context of anti–PD-1 therapy, a similar picture of PD-L1 and CD8 expression within granulomas is observed, while PD-1 immunoreactivity appears reduced (Figure 1B). Therefore, the immune checkpoint inhibition pathway is active within TB granulomas.

Figure 1.

PD-L1 and PD-1 are expressed in human lung TB granulomas. Six lung biopsies of patients with a final diagnosis of TB were immunostained for PD-L1, CD8, and PD-1. (A) PD-L1 is highly expressed by macrophages within the granuloma, and PD-1 is expressed by CD8-positive T cells. (B) In the pembrolizumab-treated patient, strong PD-L1 and CD8 immunoreactivity was observed, whereas PD-1 staining appeared less strong. Scale bars: (A) 200 μm; (B) 100 μm. PD-1 = programmed cell death-1; PD-L1 = programmed cell death ligand-1; TB = tuberculosis.

Potential Mechanisms of Immune Checkpoint Inhibition Causing TB Reactivation

PD-1 is a cell surface receptor that binds ligands PD-L1 and PD-L2 and has important functions in the maintenance of immune tolerance. PD-1 inhibitors are therefore used to reverse tolerance to tumors and improve immune-mediated control of malignant disease (2). The use of these checkpoint inhibitors has been transformative to the field (2). In TB, progression from latent to active infection is regarded as a failure of the immune response, as demonstrated by the increased incidence of TB in the context of HIV infection or after anti–tumor necrosis factor treatment for inflammatory conditions (1). Consequently, it seems highly counterintuitive that PD-1 blockade should also cause activation of TB, as by this paradigm anti–PD-1 therapy should improve host control of TB. Indeed, PD-1 inhibition has been suggested as a host-directed therapy in TB (3), on the basis that the PD-1 pathway may inhibit an effective host response.

Mechanistically, these observations suggest that immune checkpoint signaling is important to conserve immune homeostasis within TB granulomas and prevent excessive inflammation that may lead to tissue destruction and cavitation (4). In terms of the cellular events leading to TB, depletion of Mtb-responsive T cells by anti–PD-1 treatment would be most consistent with the current paradigm whereby a greater host immune response limits Mtb growth. However, this would imply a dual effect, with immune checkpoint inhibition improving control of malignancy by immune activation, while concurrently suppressing anti-mycobacterial immune responses. Therefore, an alternative process seems likely.

A rapid T cell–driven immune activation could lead to greater recruitment of permissive monocytes or neutrophils to TB granulomas, which are thought to be deleterious in TB (1). Alternatively, this augmented immunity may result in increased cytotoxicity or matrix metalloproteinase–driven extracellular matrix destruction, which favors Mtb growth and leads to transmission of infection (4). An unorthodox proposal is that active TB can result from an autoinflammatory or autoimmune process, a hypothesis supported by diverse clinical and experimental observations (5). Analysis of gene expression profiles of patients with TB, infection, and autoimmune disease also suggests a common underlying mechanism between TB and autoimmunity (6). Intriguingly, the most common adverse events from immune checkpoint inhibitors are autoimmune in nature (2), consistent with this hypothesis. Along similar conceptual lines, Divangahi and Behr have proposed that T cell–mediated immune tolerance may be equally important in host control of TB as effector functions (7).

Implications for Clinical Practice and Research

The clinical implications of this phenomenon are wide. New lesions in a patient with known cancer are likely to be diagnosed as malignant progression, leading to significant underdiagnosis. We suggest that biopsy of progressive lesions in this context is indicated to exclude TB, as it is treatable with antibiotics. Furthermore, patients with cancer should be screened by Mantoux test or IFN-γ release assay before receiving anti–PD-1 treatment, in the same way that screening is routine before anti–tumor necrosis factor treatment. If previous Mtb exposure is diagnosed, chemoprophylaxis may be indicated to prevent active TB.

Regarding the research implications, the host–pathogen interaction is finely balanced, with only a small subset of Mtb-exposed individuals developing pulmonary disease to continue transmission (1, 4). The clinical observations that are emerging from the biologic treatment era provide entirely novel and highly relevant insights into human immune function and host–pathogen interactions. The increase in active TB after immune checkpoint inhibition suggests that excessive immunity may be just has harmful as insufficient immunity. Supporting this concept, Comstock and colleagues demonstrated in a study of 82,000 individuals that a strong response to TB antigens, which should be considered protective, actually associates with progression to active TB (8). Infection of mice deficient in PD-1 results in rapidly lethal inflammation (9, 10), consistent with a protective role in TB. Perhaps the most sobering implication is that it reinforces the finely balanced knife edge of the human–Mtb interaction. Simply driving an exaggerated immune response, without first defining determinants of progression versus protection in TB, risks inadvertently accelerating transmission and worsening the pandemic in the longer term (4).