Sequential monitoring of lymphocyte subsets and of T-and-B cell neogenesis indexes to identify time-varying immunologic profiles in relation to graft-versus-host disease and relapse after allogeneic stem cell transplantation

T and B lymphocyte subsets have been not univocally associated to Graft-versus-host disease (GVHD) and relapse of hematological malignancies after stem cell transplantation (SCT). Their sequential assessment together with B and T cell neogenesis indexes has been not thoroughly analysed in relation to these changing and interrelated immunologic/clinic events yet. Lymphocyte subsets in peripheral blood (PB) and B and T cell neogenesis indexes were analysed together at different time points in a prospective study of 50 patients. Principal component analysis (PCA) was used as first step of multivariate analysis to address issues related to a high number of variables versus a relatively low number of patients. Multivariate analysis was completed by Fine-Gray proportional hazard regression model. PCA identified 3 clusters of variables (PC1-3), which correlated with acute GVHD: PC1 (pre-SCT: KRECs≥6608/ml, unswitched memory B <2.4%, CD4+TCM cells <45%; HR 0.5, p = 0.001); PC2 (at aGVHD onset: CD4+>44%, CD8+TCM cells>4%; HR 1.9, p = 0.01), and PC3 (at aGVHD onset: CD4+TEMRA<1, total Treg<4, TregEM <2 cells/μl; HR 0.5, p = 0.002). Chronic GVHD was associated with one PC (TregEM <2 cells/μl at day+28, CD8+TEMRA<43% at day+90, immature B cells<6 cells/μl and KRECs<11710/ml at day+180; HR 0.4, P = 0.001). Two PC correlated with relapse: PC1 (pre-SCT: CD4+ <269, CD4+TCM <120, total Treg <18, TregCM <8 cells/μl; HR 4.0, p = 0.02); PC2 (pre-SCT mature CD19+ >69%, switched memory CD19+ = 0 cells and KRECs<6614/ml at +90; HR 0.1, p = 0.008). All these immunologic parameters were independent indicators of chronic GVHD and relapse, also considering the possible effect of previous steroid-therapy for acute GVHD. Specific time-varying immunologic profiles were associated to GVHD and relapse. Pre-SCT host immune-microenvironment and changes of B cell homeostasis could influence GVH- and Graft-versus-Tumor reactions. The paradoxical increase of EM Treg in PB of patients with GVHD could be explained by their compartmentalization outside lymphoid tissues, which are of critical relevance for regulation of GVH reactions.

Introduction Long term efficacy of allogeneic stem cell transplantation (SCT) in haematological malignancies relies primarily on graft-versus-tumor (GVT), which partly overlaps with graft-versushost disease (GVHD) [1,2], the most common cause of morbidity and mortality in SCT [3]. However, GVT and GVHD are probably characterized by different intensity of immune reactions, which can be modulated by different subsets of donor T and B lymphocytes [1][2][3][4]. Several studies correlated T lymphocyte subtypes in peripheral blood (PB) with GVHD (acute and chronic) and relapse, although without univocal results [5][6][7][8][9][10][11][12][13][14][15][16][17][18]. The role of B lymphocytes in chronic GVHD (cGVHD) was evidenced by several authors, whereas their relationship with acute GVHD (aGVHD) and relapse has been poorly investigated [5,[19][20][21][22][23][24][25][26]. Adequate thymic function measured by quantification of T-cell receptor excision circles (TRECs) has been correlated with balanced immune reconstitution and reduced risk of infections [27][28][29]. Levels of k-deleting recombination excision circles (KRECs) have been associated with poor B lymphocyte reconstitution and cGVHD, whereas a straightforward relationship between KRECs values and aGVHD has not been evidenced yet [30][31][32]. The uncertain and controversial findings reported in literature could be partly explained by the difficulty of analysing all these immunologic variables in a high number of patients with an extensive monitoring in the time. Furthermore, most studies focused on one outcome only, GVHD or relapse, without considering their complex interplay. The rationale of our study relies on the following points: 1. T lymphocytes are the principal effectors and coordinators of immune responses, and B lymphocytes have an emerging role not only as effectors but also as long-lasting regulators of immune reactions [1,2,[19][20][21]. Hence, the importance of B and T cell neogenesis indexes as well.
2. A sequential monitoring of lymphocyte subsets and thymic and bone marrow output indexes could better fit with the changing behavior of GVHD and relapse, allowing the identification of specific immunologic indicators, which could differ depending on the time before and after SCT. In particular, the start of monitoring already before SCT could allow to identify a correlation between pre-SCT host immune profiles and GVHD or relapse. In fact, the state of host immune microenvironment at SCT, which influences alloreactions by donor lymphocytes [3,19,21], may partially depend on its state pre-SCT.
3. To our knowledge, no previous study analysed all the above-mentioned variables together before and after SCT in relation to aGVHD, cGVHD and relapse.
We prospectively evaluated T and B lymphocyte subsets together with thymic and bone marrow output indexes in 50 patients at different time points before and after SCT in relation to aGVHD, cGVHD, and relapse, as clinical indicator of ineffective GVT. We used a 2-step multivariate analysis, which included principal component analysis (PCA), to counterbalance the limits of the relatively low number of enrolled patients in comparison to the high number of variables considered in this study.

Patients and transplant procedures
Prospective evaluations of lymphocyte subsets in PB and thymic and bone marrow output indexes were performed in 50 not consecutive patients who underwent allogeneic SCT (Table 1). Acute leukaemia was the prevailing diagnosis (56%); 17 patients had lymphomas; one patient only was treated with rituximab 2 years before transplantation.
Cytomegalovirus (CMV) and HHV-6 viral replication were monitored weekly by quantitative real-time polymerase chain reaction (PCR) in plasma. Fungal infections were evaluated according to the revised criteria of EORTC/MSG Consensus Group [33]. Blood stream infection was defined according to Poutsiaka D et al. [34]. Chimerism was assessed by tandem repeats (VNTR)-PCR test on PB mononuclear cells (PBMCs), polymorphonuclear leukocytes and lymphocytes at day+30, +90, +180, and in case of disease relapse. The first evaluation of disease was performed at day +30 for acute leukaemia and myelodisplastic syndrome, and at day +60 for lymphomas. Diagnosis and grading of aGVHD and cGVHD were primarily based on clinical findings [35][36][37]. Whenever possible, clinical data were supported by histopathologic findings of target organs. cGVHD was defined as mild, moderate or severe according to NIH criteria for cGVHD grading [37]. GVHD prophylaxis included cyclosporine A (CyA) i.v., and a short course of methotrexate (MTX). ATG-Fresenius was used in 26 of patients with unrelated donor and 3 patients with related donor. CyA was given orally in two doses as soon as patients were able to have an oral intake. CyA was the only GVHD prophylaxis in all patients until its tapering, or until the beginning of the therapy against aGVHD or moderate/ severe cGVHD. First-line therapy for acute GVHD of at least grade II included CyA and 2 mg/ kg/day methylprednisone. Methylprednisone 1 mg/kg/day was the first-line therapy for cGVHD. The study was performed on a subgroup of transplanted patients (from 31/03/2011 to 01/06/2015), who were enrolled in the study "COORTE HSCT" (n.854 26/05/2010). This study was reviewed and approved by the Ethics Committee of the Hospital "Spedali Civili" of Brescia. Patients provided written consent in accordance with the Declaration of Helsinki. One minor only was included and written consent was obtained from the parents on its behalf.
From the analysis of this substudy, patients with the following characteristics were excluded or removed: • bone marrow and PB involvement by lymphoproliferative disease at SCT; • follow-up < 3 months; • absence of full chimerism at day +30 or loss of chimerism independently from relapse; • persistent disease at the first evaluation after SCT; • steroid-refractory aGVHD development.

Assessment of lymphocyte subsets and index of T and B cell neogenesis
To investigate a possible correlation of lymphocyte subsets and index of T and B cell neogenesis with aGVHD, cGVHD, and relapse, we performed: 1. a flow cytometry analysis of T and B cell subsets; 2. a quantification of TRECs and KRECs by Real-Time PCR.
The assessment was planned pre-SCT, at day +28, +90, +180 and at the time of aGVHD, which developed at a median time of 27 days (22-100). Hence, aGVHD analysis included only Lymphocyte subsets and T/B cell neogenesis indexes in graft-versus-host disease and relapse the assessment pre-SCT, at onset of aGVHD, and at day+28 for patients without it. Since cGVHD developed between the 6 th and the 10 th month (median time: 7 months), the immunologic parameters at each time point were analysed in relation to cGVHD. Relapse occurred between the 3 rd and the 12 th month (median time: 6 months), therefore the assessment of day +180 was excluded from relapse analysis.
The assessment was also performed in 15 age-matched healthy controls.
Absolute count and percentage were calculated for each T and B cell subset. Data were acquired using an eight-colour FACSCanto II cytometer and analysed with FACS Diva software (BD Biosciences, San Jose, CA, USA).

Analysis of T and B cell neogenesis by means of TRECs and KRECs quantification.
Thymic and bone marrow outputs were measured from DNA from PBMCs at different time points. KRECs and TRECs were quantified simultaneously by duplex quantitative Real-Time PCR (7500 Fast Real-Time PCR of Applied Biosystems, Foster City, CA), as described elsewhere [38,39]. Their quantities were obtained from a standard curve obtained by serial dilutions of a linearized plasmid DNA, containing three inserts corresponding to fragments of KRECs, TRECs and the reference gene, which is a fragment of T-cell receptor constant alpha gene. Data are expressed as number of copies per ml of blood, equal to (KRECs or TRECs/ PBMC) × (lymphocyte+monocyte count in 1 ml of blood).
Assessment of steroid treatment effect. In aGVHD analysis, the possible effect of steroid treatment on values of lymphocytes, TRECs, and KRECs was excluded, as they were assessed at GVHD onset before the beginning of the therapy. All patients with cGVHD were off steroid therapy at its onset. However, the influence of a previous steroid treatment for aGVHD before the occurrence of cGVHD or relapse was considered. Parameters of steroid effect were: 1. administration of steroid therapy at each time point; 2. cumulative dose of steroids (mg/kg) at day +28, from day +28 to +90, and from day +90 to +180; 3. number of days out steroids at each time point. A correlation between the parameters of steroid effect and values of lymphocyte, TRECs, and KRECs was analysed at each time point.

Statistical analysis
Univariate analysis of variables in relation to aGVHD, cGVHD and relapse was performed by Mann-Whitney U test to compare continuous values, and chi-squared test to compare differences in percentage. Count and percentage of lymphocyte subsets, TRECs and KRECs values at the different time points were analyzed together with clinical variables (see Table 1). The analysis included infectious events before the onset of aGVHD or cGVHD and before the median time of their onset for patients without these complications. aGVHD and cGVHD were also considered for relapse analysis. The comparison of lymphocyte, TRECs, and KRECs values between patients on and off steroid therapy at each time point was also performed by Mann-Whitney test. The Spearman rank correlation analysis was used to estimate the relationship between cumulative dose of steroids or days off steroid therapy and values of the immunologic variables.
The immunologic parameters found to be significant in univariate analysis at each time point were included in the first step of multivariate analysis, which consisted in PCA, in order to solve the problem of a high number of variables in comparison with a relatively limited and heterogeneous pool of patients. PCA reduces the dimensionality of a large number of interrelated variables, while retaining as much information as possible [40]. Since PCA transforms possibly correlated variables into a smaller number of uncorrelated variables (principal components, PCs), each PC is a cluster of correlated variables. The first PC (PC1) accounts for the largest part of the total variance in the dataset; the second (PC2) accounts for the second greatest amount of the variance, and so on. The last few PCs do not account for much of the variance, and therefore can be ignored. The eigenvalue-one criterion (Kaiser criterion) was used for extracting relevant PCs (eigenvalue >1). Variables with component loading > 0.5 (absolute value) only were included in each PC. Loadings vary in value from -1 to 1 and represent the degree to which each of the variables correlates with each PC.
Multivariate analysis was completed by Fine-Gray proportional hazard regression model for competing events [41,42], which included: 1. PCs scores (values of the PCs extracted by PCA for each patient); 2. clinical variables and parameters of steroid effect, which were significant in univariate analysis.
For the immunologic variables clustered in each PC, which resulted to be significant in this second step of multivariate analysis, the median value was calculated and taken as the cut point.
Death without GVHD and death without relapse were the competing events for GVHD and relapse analysis, respectively. All p values were 2-sided and p<0.05 was considered statistically significant.
Clinical and immunological characteristics of patients with aGVHD 28 patients (56%) developed aGVHD (median time: 27 days; range, 22-100); it was of at least grade II in 20 patients (grade III in 2 out of 20 patients). Before its onset, 6 patients (21%) had bacterial infections. CMV reactivations, without end-organ disease, and fungal infections were observed in 7 (25%) and 4 patients (14%), respectively. Patients with and without aGVHD did not differ in clinical and transplant characteristics (S2 Table). No significant differences were observed as far as immunosuppressive prophylaxis (use of ATG, MTX doses, CyA blood levels) and type of reduced intensity or myeloablative conditioning were concerned (data not shown). Univariate analysis of immunological variables is summarized in Table 2.
Significant differences in percentages of lymphocyte subsets are reported in the text below.

Multivariate analysis of variables associated with aGVHD
In the 2-step multivariate analysis, three PC were evidenced as independent factors associated with aGVHD, also considering the possible effect of lymphoma diagnosis and use of ATG (Table 3). PC1 was inversely correlated with aGVHD (HR 0.5; p = 0.001), and included pre-SCT values of KRECs, percentages of unswitched memory B (<2.4%) and CD4+T CM cells (<45%). PC2 was associated with aGVHD onset (HR 1.9; p = 0.01), and included percentages of CD4+ and CD8+T CM cells. PC3 was inversely correlated with aGVHD (HR 0.5; p = 0.002), and included lower values of CD4+TEM RA cells, total Treg and Treg EM .

Clinical and immunological characteristics of patients with cGVHD
The cumulative incidence of cGVHD was 26,5% (95% C.I. . It was mild in 1, moderate in 8 and severe in 4 patients according to NIH criteria for cGVHD grading.aGVHD preceded cGVHD in 11 patients (grade !II in 4 patients). Before cGVHD onset, patients had 40% bacterial infections and 20% CMV reactivation without end-organ disease. In univariate analysis, matched related donor (62% vs 28%; p = 0,03) and previous aGVHD (85% vs 44%; p = 0,01) were the only clinical variables associated to cGVHD (S3 Table). Values of immunological variables in patients with and without cGVHD are shown in Table 5. As for aGVHD analysis, differences in percentages are reported in the text.

Discussion
T and B lymphocyte subsets in PB have been investigated as potential markers of aGVHD and cGVHD or relapse without univocal results [5][6][7][8][9][10][11][12][13][14][15][16][17][18][22][23][24][25][26]. TRECs and KRECs values have  Lymphocyte subsets and T/B cell neogenesis indexes in graft-versus-host disease and relapse been mainly investigated in relation to poor immune reconstitution and risk of infections as consequence of GVHD. Their role as predictors of GVHD and relapse is not well established yet [27][28][29][30][31][32]. These controversial findings could be related to analytic approaches, which did not consider the variability over time of the complex interplay among all these immunologic variables, GVHD, and GVT together. Furthermore, a sufficiently large and homogeneous pool of patients may be difficult to obtain in the transplant setting in a relatively short time, in order to perform such a comprehensive analysis. We addressed these issues performing a prospective analysis of T and B lymphocyte subsets, TRECs and KRECs values at different time points before and after SCT, in relation to aGVHD, cGVHD, and relapse. A 2-step multivariate analysis including PCA was used to solve the problem of a relatively limited and heterogeneous pool of patients in comparison to the numerous variables. PCA has important noise-reducing properties in small populations of patients, and can achieve the same noise reduction as using large populations [40]. In this 2-step multivariate analysis, specific time-varying immunologic profiles were correlated at each outcome of the study: aGVHD, cGVHD, and relapse.  Lymphocyte subsets and T/B cell neogenesis indexes in graft-versus-host disease and relapse aGVHD was associated with an immunologic profile including pre-transplant lower KRECs values and some lymphocyte subtypes pre-SCT and at its onset. Lower KRECs values may indicate a damaged bone marrow microenvironment, predisposing to an impairment of central tolerance, and eventually to an imbalance of B cells. This may promote the prevalence of recipient B lymphocytes with pathogenic functions such as secretion of pro-inflammatory cytokines and, as antigen presenting cells (APC), activation of effector T cells or inhibition of Treg [43][44][45]. These host B-APC could survive after SCT, and stimulate donor effector T cells in GVHD reactions. Before SCT, the prevalence in PB of unswitched memory over the other mature B cells could originate from an impairment of germinal center function, as other checkpoint for B tolerance, contributing to recipient B cell imbalance [43]. Both pathogenic and protective effects of recipient B lymphocytes on aGVHD have been reported, mostly from studies performed on mice models, or from studies without distinction among B cell subtypes [19,21].
At the onset of aGVHD, patients had a higher percentage of CD8+T CM and increased values of CD4+TEM RA cells, "senescent" cells with raised levels of activation, resistant to apoptosis [46]. Their increase, never reported by other authors in aGVHD, may be the expression of dysregulated and skewed response toward host antigens, and could partly explain the resistance to aGVHD therapy. Patients with aGVHD showed also higher PB values of Treg EM . This apparently contradictory finding was rarely reported before [5,7]. Several studies have been performed in murine models of GVHD, or in transplanted patients without distinction among Treg subsets, which display different regulatory efficiency [47]. Treg EM seem less effective than naïve Treg, since naïve Treg may directly inhibit effector T cells and APC in lymphoid tissues, which are of critical relevance for induction and suppression of GVH reactions [48,49]. Unlike naïve Treg, Treg EM lack of homing receptors such as L-selectin and chemokine receptor 7 (CCR7), which are needed for migration to secondary lymphoid organs and in situ regulation of immune responses [48][49][50]. On the other hand, increased Treg EM in PB could be expression of an ineffective compensatory reaction.
The immunologic profile associated with cGVHD included Treg EM (+28), CD8+TEM RA cells (+90), and immature B cells and KRECs values (+180). Higher PB values of Treg EM at day +28 were associated with cGVHD development, highlighting their potential role as early indicator of dysregulated immune responses. The prevalence of TEM RA cells among CD8+ lymphocytes before cGVHD onset (day+90) may be another indicator of dysregulated and skewed immune system. Such dysregulation may involve the B-compartment as well. Indeed, we observed increased values of KRECs and immature B cells at day+180 in patients with cGVHD, as possible expression of hyper-stimulated B cell output, defective censoring of hostreactive B cells in bone marrow, and defective maturation to naïve B cells in secondary lymphoid organs. Increased values of KRECs in cGVHD were never reported before, while there are few studies on the impact of GVHD on B cell generation [30][31][32]. The prevalence of immature B cells in PB was evidenced in autoimmune diseases and immunodeficiences, while their association with cGVHD is controversial [22,23,25,44,45]. Immature B cells elicit rapid antibody-dependent or -independent responses in absence of T cell induction, and are not completely depleted by CD20-targeted immunotherapy [51], partly explaining the incomplete responses observed in cGVHD.
Relapse, as index of ineffective GVT, showed also its typical immunologic hallmarks, depending on the time related to SCT. Relapsed patients had an ineffective T-immune system with decreased values of CD4+, CD4+T CM , Treg, and Treg CM cells already before transplantation. We showed a possible role of different B cell subsets in relapse control both before and after SCT, while few evidences, mostly regarding antibodies production and GVT, were reported before [5,19]. The decreased percentage of recipient mature B cells in relapsed patients suggests a defect of B-immune system even before SCT.
At day+90, ineffective GVT correlated with increased values of KRECs and switched memory B cells, as possible expression of an early but ineffective B cell-hyperactivation.
Different clusters of immunological parameters at different time points were evidenced as indicators of aGVHD, cGVHD and relapse, allowing a clear-cut distinction between these immunological/clinical events. As a novel finding, we highlighted a possible role of pre-transplant host immune-microenvironment in promoting or dampening GVHD and GVT. The atypical association of Treg EM with GVHD could be explained by the different efficiency of Treg subsets. Imbalances of B-cell homeostasis appeared to be involved both in GVHD and relapse with different indicators and features, also depending on the time before or after SCT. These specific time-varying immunologic profiles could drive a targeted, time-varying modulation of both immunosuppressive prophylaxis and pre-emptive therapy. Although the 2-step multivariate analysis addresses issues related to an unfavourable ratio between number of patients and number of variables, further studies may be helpful to validate our findings.

Author Contributions
Conceptualization: CS SP LI.