T cell receptor repertoire as a novel indicator for identification and immune surveillance of patients with severe obstructive sleep apnea

Introduction

Obstructive sleep apnea (OSA), affecting up to 936 million people worldwide, of which 425 million are moderate to severe patients, is the most prevalent sleep disturbance with a high social and economic burden (Benjafield et al., 2019). Untreated patients with OSA are at higher risk (George, 2007; McNicholas et al., 2018; Bradley & Floras, 2009; Justeau et al., 2020) of metabolic syndrome and type 2 diabetes, cardiovascular and cerebrovascular morbidity, neuropsychiatric dysfunction, pulmonary hypertension, cancer, motor vehicle accidents, and even worse co-morbidity prognosis, including severe COVID-19 (Rögnvaldsson et al., 2022). Chronic inflammation is one of its most essential pathological features (Lavie, 2015), but the role of adaptive immunity and how it is altered in OSA remains unclear. Unfortunately, with inadequate access to diagnosis based on polysomnography (PSG), numerous patients with moderate to severe OSA have still not been diagnosed (McNicholas et al., 2018), especially during the pandemic.

In OSA, chronic low-grade inflammation and immune impairment are widely documented and are suspected to be responsible for systemic injury, complications, and worse prognosis (Bradley & Floras, 2009; Lavie, 2015). Studies revealed a series of changes from cytokines, chemokines, and metabolic products to blood cell categories in OSA (Yokoe et al., 2003; Ryan, Taylor & McNicholas, 2005; Cubillos-Zapata et al., 2017), as an increase in CRP, IL-6, IL-10, TNF-a, NFkB, PD-1, cell-free DNA, and neutrophil/lymphocyte ratio (NLR) (Rha et al., 2020), CD4+T/CD8+T ratio (Domagała-Kulawik et al., 2015). The macrophages population was also shifted towards the M1 subset (Khalyfa, Kheirandish-Gozal & Gozal, 2018). There is proceeding evidence ahead of the immune mechanism nature of OSA, but little is known about the initiation and linkage of the cellular immune response till now, particularly, the T cell activation and proliferation.

As is known to all, T cells are at the core of the adaptive immune in humans, responsible for infection immunity, autoimmunity, tumor immunity, and immune regulation. While its activation and proliferation depend on the combination of T cell receptor (TCR) to a corresponding MHC-antigen peptide complex (pMHC) on antigen presenting cells (APCs) (Rajewsky, 1993). Theoretically, there could be over 1 ×10¹³ possible TCR types to recognize and combat almost unlimited types of possible antigens (Nielsen & Boyd, 2018). Thus, the tremendously diversified repertoires of TCR is a snapshot of an individual’s T cell immune status, which embodies information about exposure to endogenous and exogenous antigen presented by APCs, the ability of adaptive immune system to resist new challenges, as well as the character, extent, memory of immune responses. The most variable region of TCR, the complementary determinant region 3 (CDR3), which is critical in determining antigen specificity, is generated through variant rearrangements of different V, D, and J gene segments, in addition to the template nucleotide insertion and deletion, can be a representative for each T-cell clone (Woodsworth, Castellarin & Holt, 2013). High throughput sequencing of TCR beta CDR3 alone allows to reflect directly the T-cell directory in considerable measure.

Our study, with the aim to investigate TCR repertoires alterations in peripheral blood of OSA patients and to explore potential biomarkers for severe OSA identification and assessment, found a significant increase in diversity and clonotypes in OSA patients. Further analysis of V, J gene usage and amino acid clonotypes bias confirmed the correlation with OSA severity. Then we developed a novel indicator termed OSA TCR Characteristic Index (OSA-TCI) to summarize the extent of disease-associated TCR repertoire changes, which was subsequently validated as accurate in distinguishing patients with severe OSA from others. Moreover, we observed a consistent decrease in OSA-TCI and improvement of TCR repertoire in eight severe patients after treatment during follow-up. These findings may improve our understanding of OSA’s cellular immune function impairment and provide us with a convenient peripheral blood biomarker for severe OSA assessment without long-time contact or facility/instrument occupation. In addition, it may shed light on future diagnostic, immunological, pathophysiological, and prognostic research on OSA.

Material and Methods

Results

Repertoire diversity and clonality in severe OSA

We analyzed the diversity and abundance of TCR clonotypes. The D50 diversity index and Shannon’s diversity index in both OSA groups were unexpectedly significantly higher compared to the healthy control group (Figs. 1A, 1B; Table S1), suggesting a putative increase in clonotypes in OSA patients. Furthermore, comparison of clonotypes and unique VJ counts between the healthy control and other groups confirmed that patients with OSA and other non-OSA sleep disturbances are impaired in TCR diversity and clonality (Figs. 1C, 1D; Table S1). We further performed the principal component analysis and observed that the sample plots from the OSA patients were remarkably clustered to the upper left region in the picture, significantly different from those from healthy donors (Fig. 1E).

Characteristics of V, J gene usage, and amino acid clonotypes

We then evaluated the V and J gene usage of TCRβ CDR3 repertoire in severe OSA compared to other groups (Figs. 2A, 2B, 2C; Fig. S3) and identified a series of V, J gene usage and V-J alleles that were significantly different between severe OSA and other groups (Table S2). Mild-to-moderate OSA also showed a significant difference in V, J usage and combination from Non-OSA (Fig. 2D; Fig. S4). As shown in volcano plots (Fig. 3), the expression of amino acid clonotypes differed between every two groups (Table S3). Remarkably, the clonotype of ASSSGSSYNEQF was significantly decreased in severe OSA. Moreover, the clonotype of ASGPREK showed a lower expression in both severe OSA groups and mild-to-moderate OSA.

TCR repertoire reflected therapeutic effects of nPAP in follow-up consistently

NPAP therapy is generally acknowledged as the first choice for severe OSA patients. In the follow-up, eight severe OSA patients had completed 3 months’ family non-invasive ventilation therapy of good adherence (>4 hrs per night, >20 hrs per week). We compared their TCR repertoires and observational indexes before and after treatment (Fig. S5, Table S7): Their TRBV and TRBJ usage significantly changed after nPAP (Figs. 5A, 5B). Notably, regardless of ESS level before treatment, the OSA-TCIs of all eight patients were consistently reduced and statistically different (P = 0.0078), and the values decreased greater in patients with higher OSA-TCI (Fig. 5C). Meanwhile, the D50 values and Shannon’s diversity index of treated individuals showed different degrees of increase with statistical difference (Table S7, Fig. S6). At the same time, the overall ESS also decreased after ventilator treatment, while some patients with low ESS scores did not improve further, and no significant difference was found with regard to the small sample size (Fig. 5). Additionally, we did not observe any statistical difference in NLR and CD4+/CD8+T before and after the treatment. The inconsistent trend shown by the figure could not directly reflect the effect of ventilator therapy on immune cell counts (Figs. 5E, 5F).

Discussion

Previously, numerous studies have demonstrated the relationship between OSA severity and worse clinical consequence in comorbidities (George, 2007; McNicholas et al., 2018; Bradley & Floras, 2009; Justeau et al., 2020; Rögnvaldsson et al., 2022; Lavie, 2015). Meanwhile, an increase in circulating pro-inflammatory cytokines and oxidative products such as IL-6, TNF-a, CRP, NFkB, and cell-free DNA peptides were also observed to correlate positively with AHI (Yokoe et al., 2003; Ryan, Taylor & McNicholas, 2005; Cubillos-Zapata et al., 2017; Bauça et al., 2017). Moreover, effective therapy for severe patients, such as the implementation of adherent nPAP or surgery, can not only lower the inflammatory biomarker levels (Yokoe et al., 2003; Xie et al., 2013) but also significantly improve endothelial function and microvascular blood flow (Xu et al., 2015), resistant hypertension (Iftikhar et al., 2014), sleepiness, quality of life, mood, neurocognitive function, left ventricle ejection fraction, hemoglobin A1c, fasting glucose, cardiovascular events incidence, and incident mortality (Patil et al., 2019). The inflammatory nature is therefore supposed to be one of the essential mechanisms leading to systemic injury and comorbidity deterioration in OSA (Rögnvaldsson et al., 2022; Lavie, 2015; Ryan, Taylor & McNicholas, 2005), valuable in disease surveillance and treatment. Therefore, interest focus on exploring biomarkers has been increasing for decades. In recent years, additional research on cellular immune has reported that macrophage polarization towards M1 subsets (Khalyfa, Kheirandish-Gozal & Gozal, 2018), increased ratio of NLR (Rha et al., 2020), PLR (Kıvanc et al., 2018), CD8+T/CD4+T (Domagała-Kulawik et al., 2015), and Th17/Treg (Reale et al., 2020), as well as DNA methylation of toll-like receptor 2/6 (Huang et al., 2020) in OSA patients, suggesting possible reactions to new endogenous antigens. Given the vital role of T cells in adaptive immunity, the question arises as to why and how T cell immunity is impaired in OSA.

This study provides the first investigation into TCR repertoire features in patients with OSA towards the question above. Remarkably, by sequencing, we revealed that notable disease-severity-associated clones, disease-preferred V β, J β, and V-J junctions exist in OSA and diversity and new clonotypes dramatically increased, suggesting unusual T cell activation and memory for a series of unique MHC-peptides complex exposure. By contrast, a significant decrease in diversity is a common phenomenon that can be found in most previous studies on other conditions, such as cancer, tuberculosis (Fu et al., 2019), pneumonia of COVID-19 (Chang et al., 2021), acute myocardial infarction (Zhong et al., 2019), and subarachnoid hemorrhage (Kim et al., 2020) and corresponds with activation of specific associate TCR clusters to emergent stimuli. Notably, there are several significant factors that can affect the TCR repertoire diversity: (1) aging: the thymus begins to wither after puberty, gradually losing naïve T lymphocytes production, decreasing the ability to respond to neoantigens with age (Douek et al., 1998). Our data also shows significant D50 declination with age in OSA patients and healthy donors; (2) immune response against antigens: clonal expansion after T cells activation can remarkably change the TCR repertoire diversity (Chen et al., 2017). After the elimination of antigens, the expanded TCR clones will decline, and only a tiny portion of these expanded TCR clones will be kept in the memory T cells department. (3) Death and exhaustion of T cell subsets. (4) Immune regulation: the crosstalk of immune cells, cytokines, and interferon molecular can modulate certain groups of T cell subsets. Importantly, HIF-1 is a ubiquitously expressed critical modulator in T cell survival, proliferation, and differentiation by altering the metabolic strategy of immune cells (Taylor & Colgan, 2017; McGettrick & O’Neill, 2020). Thus, impaired oxygen availability by CIH, which intensively promotes HIF-1 α production, may bring a wide range of pro-inflammatory and anti-inflammatory implications to a series of immune cells. OSA patients, featured by the unique pattern of replicated nocturnal CIH, may induce more extensive, long-lasting, and multi-elements drove adaptive alteration rather than uniformed clonal expansion induced by acute massive stimuli. Furthermore, long-term recurrent exposure to inflammatory conditions may also induce immune tolerance to maintain homeostasis, thus reducing the probability of large clones.

Of note, the adaptive immunity impairment, pre-existing inflammatory status, and frequent fluctuation of oxygen saturation overnight also mean additional danger for OSA patients during the pandemic. The reported OR for death from severe COVID-19 in OSA patients varies from 2.0 to 2.8 compared with non-OSA, when adjusted for demographic characteristics and commodities (Rögnvaldsson et al., 2022; Cariou et al., 2020), with no data for severe OSA patients of higher risk. Besides, many risk factors for OSA are shared in severe COVID-19 patients, several possible mechanisms may lead to poorer outcomes for OSA: first, the preset chronic inflammation status may predispose to an imbalanced pro-inflammatory response to infections, more susceptible to immune factor storms (Ryan, Taylor & McNicholas, 2005; Taylor & Colgan, 2017); second, the hypoxemia in severe COVID-19 patients can be further exaggerated by CIH level associated with OSA severity, leading to more severe hypoxia-reoxygenation stress injury and cascade reactions. Third, recurrent CIH leads to a high level of HIF (Taylor & Colgan, 2017; McGettrick & O’Neill, 2020), which plays a vital role in immune dysfunction, results in macrophage polarization, neutrophil apoptosis, T cell subsets alteration, as well as the TCR repertoire changes, and may impair the ability to eliminate pathogens and sustain immunity balance. Unfortunately, on the other hand, the COVID-19 pandemic has caused a worldwide dramatic shrink in accessibility for in-lab diagnosis for fear of contamination. An investigation (Grote et al., 2020) reported that compared to pre-pandemic levels, there was a significant reduction in PSG by 20% on average, with staffing levels reduced to 25% of sleep physicians and 19% for nurses and technicians, in nineteen European countries. In order to screen out substantial severe OSA patients for necessary treatment, an easy-access and accurate tool with the lowest contaminate risk is called for in urgent need.

Through machine learning of disease-specific clones, we developed a novel indicator termed OSA-TCI to metric the extent of adaptive immune dysfunction related to severe OSA. Afterward, separate sample validation identified the overall diagnostic power of OSA-TCI to a sensitivity of 90% and a specificity of 87% in distinguishing severe OSA patients from others, with an optimal cut-off of 1087, AUC = 0.914. Although it has been reported that NLR, PLR, and CD4+/CD8+T ratio can serve as a biomarker for OSA in diagnosis (Rha et al., 2020; Domagała-Kulawik et al., 2015; Kıvanc et al., 2018), these studies neither provided any necessary validation in separate samples nor exhibited the AUC of ROC; the ROC in our study failed to support their diagnostic value for the sake of their poor AUCs. Thus, there was no reliable method for diagnosing and evaluating OSA from peripheral blood before our study. As a novel approach to identifying biomarkers, TCR immune repertoires were previously proved efficient in identifying infectious diseases, autoimmune diseases, carcinoma, and even acute myocardial infarction (Fu et al., 2019; Chang et al., 2021; Zhong et al., 2019; Kim et al., 2020). The complex but unique chronic inflammatory features of OSA, including CIH-induced cell injury and apoptosis products, endothelial function impairment, abnormal metabolomics, colonized microbial flora changes, the impacts of sleep fragmentation, and even early lesion in the cardiovascular system (Lavie, 2015; Khalyfa, Kheirandish-Gozal & Gozal, 2018; Humer, Pieh & Brandmayr, 2020), can probably contribute to forming a disease-specific repertoire landscape. There are also intimate and complex interactions between CIH-induced metabolic disturbances, immune pathway shifts that affect immune cell function, and TCR repertoire successfully integrated the picture to some extent. Although immune sequencing is not economical enough currently, some scientists are exploring cheaper TCR CDR3 sequencing methods, such as Framework Region 3 Amplifi-Kation sequencing (Amoriello & Ballerini, 2020), for better utilization in a large population. Furthermore, another advantage of TCR repertoire-based assessment is that by applying previously established classifiers for a series of diseases, many conditions can be identified simultaneously from just a single blood sample (Arnaout et al., 2021).

According to the rationalization of OSA-TCI with clinical features, OSA-TCI correlates positively with AHI, R = 0.603, demonstrating that CIH may be the dominant contributor to TCR repertoire alteration. Smoking is considered to be another significant contributor, and R = 0.22 (adjusted for AHI) hints at the putative linkage between OSA-TCI and endothelial injury aside from AHI, as smoking is also an evidenced independent risk factor of OSA, endothelial function impairment, hypertension, coronary disease. In correlation analysis of PSG participants, BMI is clarified as a confounding factor that contributes a weak correlation to OSA-TCI, R = 0.342, P = 0.001, but when BMI is adjusted for AHI, R = 0.03, P = 0.772. In contrast, no significant relationship between OSA-TCI with TST, SWS sleep, REM sleep, excessive daytime sleepiness, NLR, PLR, CD4+/CD8+T ratio, or lymphocyte counts. Whereas in individuals, OSA-TCI is not exactly in a linear relationship with AHI, which may reflect the differences in hypoxia tolerance and immuno-metabolism between individuals, which can not be explained by AHI alone, but may bring about variable impacts on clinical prognosis and treatment effect. AHI, as a single biomarker calculated from one night of PSG, can reflect the frequency of respiratory events during sleep but is unable to accurately grasp the abnormalities underlying its immune, metabolic, and cardiovascular effects on each individual (Malhotra et al., 2021), and sometimes were biased by night-to-night variation or incorrect scoring. Compared to AHI, OSA-TCI is acquired from the TCR repertoire characteristic algorithm, which involves multiple individualized responses to CIH and immuno-metabolic alterations, and may have the ability to predict prognosis and therapeutic effects. To our delight, the follow-up data indicate that, in all eight severe OSA patients who accepted adherent nPAP treatment for three months (>4 hrs per day), OSA-TCI remarkably decreased compared with their baselines. With the coincident reduction of OSA-TCI, their improvement in TCR repertoire and amelioration of daytime sleepiness and fatigue suggests its potential capability for treatment assessment. Of interest, from our small sample of before-and-after data, it appears that patients with severe OSA who had higher OSA-TCI improved more with standardized nPAP therapy. Additionally, we observed the D50 value and Shannon’s Diversity index increased after treatment, for which the mechanism is not clear and may be related to the alleviation of CIH induced immune disorders, although these two indicators are not disease-specific. Before this study, only ESS was reported valuable in predicting therapeutic effects of nPAP (Malhotra et al., 2021; Marshall et al., 2006), but it seems OSA-TCI is a more sensitive biomarker, especially in some severe patients with low sleepiness scores. With an ongoing follow-up of severe OSA in the cohort, we will test our hypothesis in a larger sample amount whether patients with higher OSA-TCI may suffer a higher risk of complications and worse clinical consequences but benefit more from therapies.

Limitations of this study include no paired TCR alpha chain sequencing, no flow cytometry to sort T cell subsets, limited blood indicators, short follow-up duration, and lack of PSG on healthy controls. In light of these limitations, subsequent research should involve single-cell sequencing of circulating lymphocytes from OSA to obtain more insights. A blinded prognostic cohort is also necessary in confirming the predictive value of OSA-TCI. Currently, economic expenditure may hinder the utilization of TCR repertoire in clinical practice, therefore, we conducted direct sequencing of TCR without previous flow cytometry to make the strategy cheaper and easier to use. However, some inflammation indexes, such as CRP, IL-6, IL-10, and TNF- α, were not considered a potential biomarker for OSA in our study for their poor specificity.

Above all, our study unveiled that the TCR repertoire changes remarkably in OSA patients, especially in severe patients, suggesting that an underlying auto-immune mechanism is closely related to intermittent hypoxia, smoking and proposed a convenient indicator for disease identification and surveillance without long-time contact and facility/instrument occupation. Furthermore, as an essential aspect of immune dysfunction in OSA, the TCR repertoire may be helpful in future therapeutic monitoring, complication risk prediction, and immunological and pathophysiological study.

Supplemental Information