Imbalance of Lymphocyte Subsets and CD45RA-Expressing Cells in Intrathoracic Lymph Nodes, Alveolar Compartment and Bloodstream of Pulmonary Sarcoidosis Patients

Sarcoidosis is a systemic granulomatous disease mainly affecting the lungs and hilomediastinal lymph nodes. It is characterized by non-caseating epithelioid cell granulomas in lymph nodes and lungs. Our study aimed to evaluate and compare T, B and NK cell subsets in the alveolar compartment, lymph nodes and the bloodstream simultaneously in the same patients to elucidate the immune responses associated with the development and progression of sarcoidosis. A secondary aim was to evaluate the distribution of CD45RA-expressing cells in the different anatomical compartments. Patients suspected to have sarcoidosis and who underwent bronchoscopy with bronchoalveolar lavage (BAL), lung-draining lymph node (LLN) biopsy by EBUS-TBNA and peripheral blood (PB) sampling were included in the study. They were monitored at the Regional Referral Centre of Siena University Hospital and the Respiratory Diseases Unit of Perugia Hospital. Multicolour flow cytometry analysis through FASCLyric was performed to assess T, B and NK cell subsets. Thirty-two patients (median age (IQR) 57 (52–58) years) were consecutively and prospectively enrolled. Machine learning analysis created a model which selected CD56dim16bright, CD8, Tfc, Th17, Th12, Tfh17, Tfh2, TcemRA, ThemRA, T naïve, Tc naïve, Breg, CD1d+CD5+, Th-reg, Tfh, Th1 and CD4 cells with an accuracy of 0.9500 (kappa 0.8750). Comparative analysis found 18 cell populations that differed significantly between the three anatomical compartments. The bloodstream was enriched in ThemRA (p = 0.0416), Tfh2 (p = 0.0189), Tfh17 (p = 0.0257), Th2 (p = 0.0212), Th17 (p = 0.0177), Th-naïve (p = 0.0368), CD56dimCD16bright (p < 0.0001), CD8 (p = 0.0319), TcemRA (p < 0.0001) and Tfc cells (p = 0.0004) compared with the alveolar compartment, while Th-reg were lower in PB than BAL (p = 0.0329). The alveolar compartment was enriched in Breg (p = 0.0249) and CD1d+CD5+ (p = 0.0013) with respect to LLN samples and PB. Conversely, Tfh (p = 0.0470), Th1 (p = 0.0322), CD4 (p = 0.0486) and Tc-naïve (p = 0.0009) were more abundant in LLN than in BAL and PB. It has been speculated that changes in the relative contents of PB cells could be related to changes in production and to the selective redistribution of PB cells to granulomatous foci. This study further supports the fact that sarcoidosis is multisystemic in nature. However, the low level of immune cells in peripheral blood of patients with sarcoidosis is concerning. A re-expression of CD45RA on CD4+ and CD8+ cells could result in a reduction in peripheral immune activity. Thus, changes in the spectrum of the bloodstream may reflect both pathogenic and compensatory processes.


Introduction
Sarcoidosis is a systemic granulomatous interstitial lung disease (ILD) mainly affecting the lungs and hilomediastinal lymph nodes [1]. It is characterized by non-caseating epithelioid cell granulomas in lymph nodes and lungs [2].
Since the disease has a typical constellation of clinical and chest X-ray findings, a presumptive diagnosis can often be advanced on the basis of a finding of radiological stage I (bilateral hilar and/or mediastinal lymphadenopathy) or stage II (bilateral hilar adenopathy and parenchymal infiltration) associated with specific clinical symptoms [3]. However, in most cases cytohistological confirmation of granulomatous inflammation is still necessary for diagnostic purposes and to exclude malignancy and infection. Sarcoid lung involvement can be evaluated by flexible bronchoscopy with bronchoalveolar lavage (BAL), cytological analysis and evaluation of T cell expression, typically characterized by an increase in alveolar lymphocytes having an elevated CD4/CD8 ratio [4]. For better diagnostic accuracy, it is advisable to perform endobronchial ultrasound-transbronchial needle aspiration (EBUS-TBNA) during flexible bronchoscopy, a minimally invasive technique that allows sampling of lung-draining lymph nodes (LLNs) for cytological analysis [5,6]. This procedure has been widely adopted and is currently recommended by international guidelines as the gold standard for evaluating intrathoracic lymphadenopathy in cases of suspected sarcoidosis [7].
Originally described as a Th1-driven disease, sarcoidosis involves a complex interplay of different immune cells. Cumulative evidence suggests that Th17, regulatory T helper (Th-reg) and follicular T helper (Tfh) cells also play a critical role in the pathogenesis [8][9][10][11]. As a result of the defective protective function of Th-reg cells in patients with autoimmune diseases, tissue tolerance is broken down, and ongoing immune responses do not decrease in a timely manner, as is the case in patients with inflammatory diseases [12].
Evidence that Th17 cells are elevated in the lung and peripheral blood of patients with active sarcoidosis sustains the multisystem nature of the disease but raises the issue of a diminished cell immune response in the blood. Both sarcoidosis and autoimmune diseases showed an activation of innate immunity that led to T reg cell dysfunction and elevated Th-17 cell count in peripheral blood. Indeed, a significant increase in the Th17/Th-reg cell ratio has been observed in BAL and peripheral blood of sarcoidosis patients with active disease [13]. These cell subsets are closely involved in the development of autoimmune diseases. Although sarcoidosis is not considered an autoimmune disease, various triggers have been implicated in its development, including autoantigen-specific T cells, antibodies producing B lymphocytes and autoimmune inflammation [14].
Several triggers including viral infections, inflammation and ageing could induce the accumulation of highly differentiated CD45RA re-expressing memory T cells [15]. There is a decrease in proliferative capacity, an increase in senescence signaling pathways activation, and a greater susceptibility to apoptosis in these cells [16]. No studies reported their role and distribution in sarcoidosis patients.
Previously neglected cells, including B lymphocytes and natural killer (NK) cells, have recently been studied to clarify their contribution to sarcoid granuloma formation [4,17]. Imbalance in B cell subsets could be linked to the abnormal distribution of Tfh cells, known to have distinct capacities to help B cells [18]. When NK and T cell subsets were recently characterized, sarcoidosis patients showed a distinct phenotype of these subsets in the bloodstream and lungs [19].
To our knowledge, a detailed comparison of their simultaneous distribution in three anatomical compartments is still lacking. Data on the distribution of T, B and NK lymphocyte subpopulations could help clarify immune response activity by defining the way certain functional cell populations are recruited to the site of damage.
Our study aimed to evaluate and compare T, B and NK cell subsets in BAL, LLN and peripheral blood (PB) simultaneously in the same patients in order to elucidate aspects of the immune responses associated with the development and progression of sarcoidosis.
A secondary aim was to evaluate the distribution of CD45RA-expressing cells in the different anatomical compartments.

Study Population
Thirty-two patients (median age (IQR), 57 (52-58) years) were consecutively and prospectively enrolled in the study. Clinical and immunological findings are reported in Table 1. Nineteen patients (59%) were females and 21 (66%) were never-smokers. At onset, 12 patients (37%) reported clinical symptoms including cough, dyspnoea, chest pain and joint pain. All patients were in Scadding radiological stage II with enlargement of lymph nodes and interstitial lung abnormalities. None of them were taking oral steroids or inhaled therapy at the time of sampling. BAL cell pattern showed lymphocytosis with elevated CD4/CD8 ratios.

Multivariate Analysis
Machine learning analysis with variable importance plot was used to select variables: it indicates how much a given model "uses" that variable to make accurate predictions. The more a model relies on a variable to make predictions, the more important it is for the model. The model ( Figure 1) selected CD56dim16bright, CD8, Tfc, Th17, Th12, Tfh17, Tfh2, Tc em RA, Th em RA, T naïve, Tc naïve, Breg, CD1d + CD5 + , Th-reg, Tfh, Th1 and CD4 cells with an accuracy of 0.9500 (kappa 0.8750).
The supervised Principal Component Analysis plot ( Figure 2) shows how the three groups (PB, LLN, BAL) separated on the basis of selected variables. The first and second components explained 44.6% of the total variance.
The alveolar compartment was enriched in Breg (p = 0.0249) and CD1d + CD5 + (p = 0.0013) with respect to LLN samples and PB. Conversely, Tfh (p = 0.0470), Th1 (p = 0.0322), CD4 (p = 0.0486) and Tc-naïve (p = 0.0009) were more abundant in LLN than in BAL and PB. NK-, Tc-and Tfc-subsets were detected in few LLN samples and were not included in the statistical analysis comparing the three compartments.

Discussion
Here we analysed the percentages of T-, B-and NK-cell subsets in three anatomical compartments of sarcoidosis patients: alveolar, lymph nodes and peripheral blood. The liquid-based cytology test in intrathoracic lymph nodes sampled by EBUS-TBNA was performed by flow cytometry, identifying the cell subsets involved in granuloma formation.
Machine learning analysis allowed us to construct a model that selected CD56 dim 16 bright , CD8, Tfc, Th17, Th12, Tfh17, Tfh2, Tc em RA, Th em RA, T naïve, Tc naïve, Breg, CD1d + CD5 + , Tfc17.1, Th-reg Tfh, Th1 and CD4 cells to make accurate predictions and variables that could be dropped from the model because they did not contribute much information. Multivariate analysis of the data reduced the number of dimensions and a total variance of 44.6% was obtained based on T, B and NK cell subsets showing good clustering variables for LLNs, BAL and PB.
Our current knowledge of the pathogenesis of sarcoidosis includes sequential immunological events triggering activation of T and B lymphocytes. This eventually leads to their migration to inflammatory foci and the formation of granulomas [20]. The present study confirmed data in the literature on depletion of CD56 dim 16 bright and CD8 + T cells in LLN and BAL samples with respect to PB, suggesting that these cells play a protective role in sarcoidosis [5,17].
In line with previous reports, our cohort showed significantly lower percentages of CD4 + T cells in PB than in LLN, further substantiating the crucial role of Th-cell interaction with antigen-presenting cells for granuloma formation and maintenance [6]. Moreover, our study is the first to describe altered Th-CD45RA − subsets in three anatomical compartments of sarcoidosis patients with higher Th2 and Th17 cell percentages in PB than in BAL samples, and higher percentages of Th1 and Treg in LLN and BAL samples than in PB, respectively. Besides supporting the multisystem nature of the disease [21,22], these findings raise concerns about low cell immunity in peripheral blood of sarcoidosis patients [23]. The regulatory counterpart of Th cells accumulating in the alveolar compartment may be due to peripheral recruitment; nevertheless, the proportion of regulatory CD127-expressing cells (Th naïve) was lower in the alveolar compartment than in PB, probably because pathogenic T cells are resistant to suppression signalling by Treg cells [24]. Assessment of CD127 expression on CD8-positive cells (Tc naïve) showed that Tc naïve cell percentages were lower in BAL than in PB and LLN.
Concerning B cells, our study is the first to report a higher percentage of CD5expressing cells in BAL than in PB and LLN samples. However, the role of this finding remains unclear in sarcoidosis. In many autoimmune disorders, CD5 + B cells are reported to expand and to be a major source of natural polyreactive and autoreactive autoantibodies [25]. CD5 + B cells must therefore contain a wealth of information about the autoimmune system that has yet to be discovered and analysed [26]. The development of therapeutic strategies to downregulate CD5 + B cells so as to mitigate ongoing autoimmunity and protect high risk individuals could be achieved if it turns out that autoimmune repertoires are concentrated in CD5 + B cells, the latter being readily identified by flow cytometry.
An imbalance in B cells may be related to the abnormal distribution of Tfh cells, found in secondary lymphoid organs, including lymph nodes, and known to have different functions, including producing plasma and memory B cells through germinal centre reactions [8,18]. Interestingly, it has been reported that Tfh cells infiltrate sarcoid skin lesions, suggesting that they are recruited to affected tissues [27]. In line with this, our study found higher percentages of Tfh cells in LLN samples than in PB and BAL of sarcoidosis patients. An abnormal distribution of Tfh-CD45RA − subsets has been linked to the pathogenesis of sarcoidosis [8,27]. Although all the lymphocyte populations mentioned above are functional cells, as far as non-expression of CD45RA is concerned, a portion of cells re-expresses CD45RA. In our study, the percentages of Th em RA and Tc em RA cells were higher in PB than in BAL or LLN. Re-expression of CD45RA on CD4 + and CD8 + cells could mean lower immune activity at peripheral level [28,29] and higher activity in the organs targeted by the disease (alveolar and lymph node compartments). Thus, inflammation is localised at the site of damage, which recruits cells from the peripheral blood where they decrease in favour of the target organ.
The present study is the first to compare the distribution of 'regulatory' Tfh1 and 'proinflammatory' Tfh2 and Tfh17 subsets in three anatomical compartments at the same time from sarcoidosis patients. Like Th2 and Th17, higher percentages of Tfh2 and Tfh17 cells were recorded in PB than in BAL samples. A hallmark of systemic autoimmune disease, as well of sarcoidosis, is the predominance of the Tfh17 subset over Tfh1 cells [9,30]. These findings suggest that targeting Tfh cells could be a therapeutic strategy for sarcoidosis, like the highly effective blockade of the main ligand for CXCR5 receptors in experimental models of rheumatoid arthritis and multiple sclerosis.
The present results improve our understanding of the immunological pathways involved in the pathogenesis of sarcoidosis and warrant a larger cohort study. One limit of our study was that healthy controls could not be included for ethical reasons.
There has been speculation that changes in the relative content of peripheral blood cells may be due to changes in cell production and to selective redistribution to granulomatous foci. This implies that any changes to the spectrum of peripheral blood cells may reflect both pathogenic and systemic compensatory processes.

Study Population
Patients suspected to have sarcoidosis and monitored at the Regional Referral Centre of Siena University Hospital and Respiratory Diseases Unit of Perugia Hospital were consecutively recruited for the study. All patients underwent bronchoscopy with BAL, lung-draining lymph node (LLN) biopsy by EBUS-TBNA and peripheral blood sampling. Diagnosis of sarcoidosis was made by multidisciplinary evaluation of clinical features, radiological features and non-caseating granulomas in lymph nodes and/or endobronchial biopsy specimens according to international ATS/ERS/WASOG criteria [31,32]. Patients with lymph node calcifications, or whose PB, BAL and LLN samples were not collected simultaneously, were excluded from the study. Prior to bronchoscopy and biopsy (by EBUS-TBNA reported in Supplementary Materials Table S1), high resolution computed tomography of the chest revealed lymph node enlargement in all patients included in the study.
Written informed consent to participate in the study was obtained from all patients. Healthy controls were not enrolled for ethical reasons. The study was approved by the regional ethics committee of Siena, Italy (C.E.A.V.S.E. Markerlung 17431 approved on 15 June 2020).

Single Cell Preparations from Blood, Bronchoalveolar Lavage Fluid and Lymph Nodes
Peripheral blood mononuclear cells (PBMCs) were isolated from blood collected in EDTA tubes according to the manufacturer's protocol [33]. BAL samples were kept on ice, filtered through a 100 µm nylon filter (Syntab) and centrifuged at 400× g for 15 min. LLN aspirates were filtered through a 40 µm nylon cell strainer and centrifuged at 300× g for 10 min. Lysis of red blood was performed with 1× BD FACS lysing solution (BD Biosciences, San Jose, CA, USA) for 5 min; the cells were then centrifuged at 300× g for 5 min. Cells were counted manually and Trypan Blue was used to assess viability as previously reported. All samples showed at least 95% cell viability.

Flow Cytometry Gating Strategy
Multicolour flow cytometry analysis was performed to assess T, B and NK cell subsets using the mAbs reported in Table 2. The cell subsets were characterized by FASCLyric (BD Biosciences, CA, USA) and they were summarized in Table S1. The gating strategy performed by Kaluza Software 2.1 (Beckman Coulter, Brea, CA, USA). For the analysis of cells, the total NK cell population was identified based on FSC vs SSC and negative for CD3, CD14, and CD19. CD56 was plotted against CD16 to obtain immature (CD56 bright CD16 neg ) and mature (CD56 dim/neg CD16 bright ) phenotypes of NK cells (Figure 3a).

PerCPCy5 BD
The gating strategy performed by Kaluza Software 2.1 (Beckman Coulter, Brea, CA, USA). For the analysis of cells, the total NK cell population was identified based on FSC vs SSC and negative for CD3, CD14, and CD19. CD56 was plotted against CD16 to obtain immature (CD56 bright CD16 neg ) and mature (CD56 dim/neg CD16 bright ) phenotypes of NK cells (Figure 3a).
The B cells was identified based on population that express CD45 + APC/CD19 + PECy7. To obtain B regulatory cells CD27 was plotted against CD24, while the population CD38 + /CD24 + was identify as immature B cells. Regulatory B cells was identified as CD1 + /CD5 + (Figure 3b). T tubes was assessed for lymphocytes discriminated on the basis of the forward (FCS) versus side (SSC) scatter. A dot plot was performed to distinguish CD8 from CD4 expressing cells. Using CD45RA as marker was identify two CD4 subpopulation CD4 + 45RA − (Thcm) and CD4 + 45RA + (Them). From CD4 + 45RA − that expressed CCR6 were identify the T cells helper 17 (CCR4 + CXCR3 − ) and T cells helper 17.1 (CCR4 − CXCR3 + ). A dot plot was performed for CD4 + 45RA − CCR6 − to discriminate T cells helper 2 (CCR4 + CXCR3 − ) and T cells helper 1 (CCR4 − CXCR3 + ). In another dot plot the CD4 expressing CXCR5 was identify as T follicular helper cells, then discriminated the Tfh The B cells was identified based on population that express CD45 + APC/CD19 + PECy7. To obtain B regulatory cells CD27 was plotted against CD24, while the population CD38 + /CD24 + was identify as immature B cells. Regulatory B cells was identified as CD1 + /CD5 + (Figure 3b).

Statistical Analysis
All data is reported as median and interquartile range (IQR) or mean ± standard deviation, as appropriate. The Shapiro-Wilk test was used to determine normal distribution. Multiple comparisons were assessed by non-parametric one-way ANOVA (Kruskal-Wallis test) and the Dunn test. Correlations between immunological and clinical findings were investigated by Spearman's test. Machine learning and variable importance analysis were performed to construct a model to select variables for accurate predictions based on the cell subsets identified in BAL, LLN and PB. Supervised Principal Component Analysis was performed in an explorative approach to identify trends in immunological features by 2D representation of the multi-dimensional data set. A p-value less than 0.05 was considered statistically significant. Statistical analysis was performed with GraphPad Prism 9.3 and Jamovi software 2.3.

Statistical Analysis
All data is reported as median and interquartile range (IQR) or mean ± standard deviation, as appropriate. The Shapiro-Wilk test was used to determine normal distribution. Multiple comparisons were assessed by non-parametric one-way ANOVA (Kruskal-Wallis test) and the Dunn test. Correlations between immunological and clinical findings were investigated by Spearman's test. Machine learning and variable importance analysis were performed to construct a model to select variables for accurate predictions based on the cell subsets identified in BAL, LLN and PB. Supervised Principal Component Analysis was performed in an explorative approach to identify trends in immunological features by 2D representation of the multi-dimensional data set. A p-value less than 0.05 was considered statistically significant. Statistical analysis was performed with GraphPad Prism 9.3 and Jamovi software 2.3.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.