Baseline T-lymphocyte subset absolute counts can predict both outcome and severity in SARS-CoV-2 infected patients: a single center study

The aim of this study was to evaluate the role of baseline lymphocyte subset counts in predicting the outcome and severity of COVID-19 patients. Hospitalized patients confirmed to be infected with Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) were included and classified according to in-hospital mortality (survivors/nonsurvivors) and the maximal oxygen support/ventilation supply required (nonsevere/severe). Demographics, clinical and laboratory data, and peripheral blood lymphocyte subsets were retrospectively analyzed. Overall, 160 patients were retrospectively included in the study. T-lymphocyte subset (total CD3+, CD3+ CD4+, CD3+ CD8+, CD3+ CD4+ CD8+ double positive [DP] and CD3+ CD4− CD8− double negative [DN]) absolute counts were decreased in nonsurvivors and in patients with severe disease compared to survivors and nonsevere patients (p < 0.001). Multivariable logistic regression analysis showed that absolute counts of CD3+ T-lymphocytes < 524 cells/µl, CD3+ CD4+ < 369 cells/µl, and the number of T-lymphocyte subsets below the cutoff (T-lymphocyte subset index [TLSI]) were independent predictors of in-hospital mortality. Baseline T-lymphocyte subset counts and TLSI were also predictive of disease severity (CD3+  < 733 cells/µl; CD3+ CD4+ < 426 cells/µl; CD3+ CD8+ < 262 cells/µl; CD3+ DP < 4.5 cells/µl; CD3+ DN < 18.5 cells/µl). The evaluation of peripheral T-lymphocyte absolute counts in the early stages of COVID-19 might represent a useful tool for identifying patients at increased risk of unfavorable outcomes.

Statistical analysis. Differences between groups were assessed using the Mann-Whitney U test, Kruskal-Wallis test (continuous variable) or Chi 2 test (categorical variables), as appropriate. Linear correlation was assessed using the Spearman's correlation test. Univariable and multivariable regression analyses were performed. Cutoff values to differentiate between survivors and nonsurvivors, as well as nonsevere and severe patients, were identified with the receiver operating characteristic (ROC) analysis and confirmed by the Youden's index. Statistics were performed using JASP (Version 0.11.1. JASP Team, 2019) and Prism 8 for macOS (version 8.2.1. GraphPad Software, San Diego, California USA). A two-sided p value of < 0.05 was considered statistically significant.
The time from symptom onset to the first NPh swab and the first cytofluorimetric assessment was similar between survivors and nonsurvivors, eliminating important confounders, such as diagnosis or hospital admission delay, that did not affect the differences observed in laboratory findings. The interval between NPh and cytofluorimetric assessment was due to the different settings in which the exams were performed: NPh swab in the emergency room and lymphocyte subset characterization upon admission to the infectious disease ward.

Results
Study population. One hundred sixty-four consecutive patients hospitalized for COVID-19 from March 8th to May 7th 2020 were enrolled. Four patients were excluded due to missing baseline information (Fig. 1). All patients had at least one confirmed positive molecular test for SARS-CoV-2 RNA detection on a NPh swab. The main clinical condition on admission was SARS-CoV-2-related pneumonia, accounting for 94.4% of the patients (151/160). For the remaining nine patients, hospitalization was due to neurological symptoms (n = 2), fever with immunocompromising condition (n = 2), exanthema (n = 1), or fever with the need for isolation due to public health reasons (n = 4) (Fig. 1). Assessment of peripheral blood lymphocyte subsets was available at baseline for all the included patients. The median age was 62 years, with a prevalence of males (61.2%). The median time from symptom onset to the first positive SARS-CoV-2 NPh swab was 4 days, while the time from symptom onset to lymphocyte subset assessment was 7 days. The majority of the enrolled population (86.3%, 138/160) had at least one comorbidity. Specifically, 50.6% (81/160) had cardiovascular diseases (mainly hypertension), 21.3% (34/160) had neurological/psychiatric disorders (mostly age-related neurocognitive impairment), 16 16.9% (27/160). The in-hospital mortality rate (nonsurvivors) was 21.3% (34/160) ( Table 1). After stratifying patients according to in-hospital mortality, older age (p < 0.001), ICU admission (p < 0.001), the presence of at least one comorbidity (p = 0.01), cardiovascular comorbidities (p < 0.001), diabetes (p = 0.029), pulmonary comorbidities (p = 0.006), renal impairment (p < 0.001) and neurological/psychiatric disorders (p = 0.026) were associated with in-hospital mortality, at the univariate analysis. Patients with a negative outcome were also more frequently treated with high flux oxygen (NRM) and noninvasive or invasive ventilation (NIV, OTI) compared to survivors (p < 0.001) ( Table 1). At the multiple logistic regression analysis, which included age, gender and comorbidities, older age, male sex and pulmonary and renal disease were independent predictors of increased in-hospital mortality (Supplementary Table 2). Flow cytometry findings. Baseline relative and absolute counts of circulating B-, T-and NK-lymphocytes were determined. After stratifying patients into survivors and nonsurvivors according to outcome (in-hospital mortality), the latter showed decreased relative and absolute counts of total CD3+ (p < 0.001), CD3 + CD8+ (p < 0.03 and p < 0.001, respectively), CD3 + CD4 + CD8+ double positive (DP) (p = 0.04 and p < 0.001, respectively) and CD3 + CD4 − CD8− double negative (DN) (p < 0.001) T-lymphocytes (Table 2). For CD3 + CD4+ T-lymphocytes, only the absolute counts were significantly reduced in nonsurvivors compared to survivors (p < 0.001). No differences were found in the CD4/CD8 ratio between the two groups. For B-lymphocytes, although the percentage of CD19+ cells was significantly increased in nonsurvivors (p = 0.03), no significant differences were seen in the absolute count between the two groups. NK absolute counts were reduced in nonsurvivors, although at the limit of statistical significance (p = 0.05) ( Table 2).
Given the greater clinical relevance of lymphocyte subset absolute count differences, compared to percentage differentials, further analyses were performed on absolute counts only.

Discussion
We retrospectively analyzed the clinical and laboratory parameters of 160 patients with confirmed SARS-CoV-2 infection and showed how T-lymphocyte baseline absolute counts help to predict COVID-19 severity and inhospital mortality. As reported in the literature, older age, male sex, the presence of comorbidities, specifically involving pulmonary and renal systems, were independent factors statistically associated with COVID-19 in-hospital mortality [10][11][12] .
Several studies have shown that lymphopenia is a peculiar finding in SARS-CoV-2-infected patients, even in the early stages of the disease 13,14 . Lymphopenia and N/L ratio have been associated with disease severity, thus allowing for their use as predictive markers of increased risk of in-hospital mortality and ICU admission 8,15 . Our study confirmed these previous results, showing that the baseline lymphocyte absolute count was reduced and the N/L ratio increased in nonsurvivors compared to survivors. We additionally analyzed lymphocyte subsets with more accuracy, demonstrating that the lymphocyte baseline reduction essentially affects the T cell compartment, while B and NK-cells seem to be less influenced by SARS-CoV-2 infection and are not related to disease severity. Interestingly, the baseline reduction of T-lymphocyte absolute count was more pronounced in patients with a more severe disease course, being independent predictors of disease severity. Previous studies have already described the potential role of T-lymphocyte subset absolute counts at baseline to assess the risk of death, progression towards a more severe disease and ICU admission among COVID-19 patients [16][17][18][19] . The majority of these studies were conducted in China during the first phase of the COVID-19 pandemic. Studies involving patients from different geographical areas and later during the pandemic are needed to confirm and extend the results from previous works. Baseline T-lymphocyte subset absolute counts were predictive of mortality, while baseline B and NK-cell absolute counts were not associated with COVID-19 mortality. Due to the concomitant reduction of CD4+ and CD8+ T-lymphocyte absolute counts, the CD4/CD8 ratio was not predictive of mortality. He et al. dynamically evaluated peripheral blood lymphocyte subset absolute counts and demonstrated that the nadir of CD4+ and CD8+ T-lymphocytes can persist for several weeks after symptom onset in severe and fatal cases, while a gradual reconstitution is observed in patients who recover 20 . These aspects further confirm the      www.nature.com/scientificreports/ potential application of T-lymphocyte subset absolute counts as a marker of disease severity and fatal outcome, even later during the disease course.
To the best of our knowledge, this is the first study in which CD3 + CD4 + CD8 + DP and CD3 + CD4 − CD8 − DN T-lymphocyte subsets have been considered for clinical purposes. CD3 + CD4 + CD8 + DP T-lymphocytes have been regarded as T-lymphocyte premature precursors, which can be released from the thymus and continue their maturation in the peripheral blood. However, increasing evidence has demonstrated that this subset of DP T-cells represents mature effector memory T-lymphocytes with a T-helper 1/T-cytotoxic 1 profile, and its absolute counts are increased in the peripheral blood of individuals with chronic viral infections 21 . Baseline CD3 + CD4 + CD8 + DP T-lymphocyte absolute counts progressively decreased in patients with more severe COVID-19 disease.
Peripheral CD3 + CD4 − CD8 − DN T-cells are considered regulatory T-lymphocytes, both naïve and antigenexperienced, representing approximately 1-5% of total CD3+ T-cells 22,23 . The role of these cells is still debated, and they have been associated with autoimmune diseases, parasitic and viral infections 24 , such as HIV, where increased absolute counts have been observed in patients with high viral loads 25 . The regulatory effect exerted by this T-lymphocyte subset can also contribute to downregulation of immune-activation through the production of immune modulating cytokines, such as transforming growth factor-beta and IL-10 26 . In our cohort, CD3 + CD4 − CD8 − DN T-lymphocyte absolute counts were significantly reduced in nonsurvivors and progressively decreased from mild to severe COVID-19 patients.
The reduction of peripheral lymphocytes, particularly of T-cell subsets, in COVID-19 patients is still not completely understood. Different phenomena, not mutually exclusive, might be involved: (1) direct infection of T-lymphocyte by SARS-CoV-2 (through the cellular receptor basigin [BSG/CD147]) and consequent cytolysis or apoptosis 27 ; (2) lymphocyte migration in the lungs, as assessed by histochemical studies from fatal cases 28,29 ; and (3) exhaustion of circulating T-cells of severe COVID-19 patients, characterized by reduced replicative abilities upon stimulation 30,31 . Taken together all these elements can contribute to lymphocyte reduction in the peripheral blood of SARS-CoV-2 infected patients with severe disease.
ROC analysis led to the identification of cutoff values for absolute counts of total CD3+ and T-lymphocyte subsets, which were predictive of disease severity and in-hospital mortality. Our results were similar to those obtained by other authors in different clinical cohorts 20,32,33 . The consistency and comparability of T-lymphocyte subset cutoffs corroborate their utility in clinical practice as prognostic and risk assessment markers in COVID-19 patients. Moreover, the TLSI can represent a useful index with the ability to summarize alterations of different T-lymphocyte subsets, indicating that a higher value of the index is associated with an increased risk of severe disease and in-hospital mortality.
In COVID-19 patients, plasma concentrations of several proinflammatory cytokines and chemokines are extremely increased, characterizing the so-called "cytokine storm", and they have been associated with disease severity and unfavorable outcomes 3 . In our cohort, we confirmed that baseline IL-6 levels were increased in nonsurvivors compared to survivors and in patients needing high flux oxygen or mechanical ventilation compared to patients with a milder disease course. No differences were found in TNF-alpha plasma levels. Interestingly, serum levels of IL-6 and other inflammatory markers, such as CRP and D-dimers, but not TNF-alpha, were inversely correlated with T-lymphocyte subset absolute counts. Although the p values of the Spearman's test were highly significant, r coefficients were < 0.5, indicating a weak correlation. Nevertheless, it is relevant that only T-lymphocyte subsets were inversely correlated to those inflammation markers which seem to play a role in the immunopathogenesis of SARS-CoV-2 infection, such as IL-6. These findings confirm the results obtained by other groups in previous works 17,31,33 .
Regarding laboratory baseline parameters, some (WBC and neutrophil counts, N/L ratio, CRP and IL-6 values) showed the most marked alterations in the NRM group, and not in the NIV or OTI groups, as expected. Furthermore, patients in the NRM group showed median values for T-lymphocyte subsets similar or even inferior to the NIV or OTI groups. These aspects could be explained by considering that, although some patients in the Table 7. Univariable and multivariable logistic regression analysis for severity-related risks in patients with SARS-CoV-2 infection. Statistically significant p values are highlited in bold. N/L ratio neutrophil-tolymphocyte ratio, CRP C-reactive protein, LDH lactate dehydrogenase, AST aspartate transaminase, INR international normalized ratio, IL-6 interleukin-6. Univariable and multivariable logistic regression were performed. Odds ratios and 95% Confidence Interval (CI 95%) are reported. A two-sided p value of < 0.05 was considered statistically significant. a Adjusting for age, gender, number of comorbidities, hemoglobin, white blood cell absolute counts, N/L ratio, CRP, LDH, AST, Creatinine, INR, d-dimer and IL-6. b The number of T-lymphocyte subset absolute counts under the cut-off value, ranging from 0 to 4.  www.nature.com/scientificreports/ NRM group were candidates for mechanical ventilation, either the presence of contraindications or the refusal to allow the procedure led to NRM being continued. The limitations of this study are its retrospective nature, the single-center design and the absence of a validation prospective cohort. Further studies are needed to confirm our results.

Conclusion
In conclusion, the most relevant finding of our study is that T-lymphocyte subsets assessed at hospital admission in a standardized and reproducible protocol, are reduced in patients with increased risk of disease progression and unfavorable outcomes.
In the present study, CD3 + CD4 + CD8 + DP and CD3 + CD4 − CD8 − DN "nonconventional" T-lymphocyte subset absolute counts were considered for clinical purpose. Furthermore, the number of T-lymphocyte subset below the cutoff-value (TLSI) was calculated for each patient and considered for predicting disease severity and in-hospital mortality of COVID-19 hospitalized patients.
Specifically, total CD3+ T-lymphocyte, CD3 + CD4+ subset absolute counts and the TLSI were independent predictors of in-hospital mortality, together with older age, male gender, increased LDH and creatinine plasmatic levels, in COVID-19 hospitalized patients.
Total CD3+, all T-lymphocyte subset absolute counts and the TLSI were independent predictors of disease severity in COVID-19 hospitalized patients.
Hence, the assessment of baseline T-lymphocyte subset absolute counts might represent a useful tool for identifying patients with increased risk of disease progression and unfavorable outcomes.
In a global pandemic scenario with limited resources, the possibility of stratifying patients early based on disease severity and outcome risk factors represents a pivotal tool to allocate resources.

Data availability
The data supporting this study will be made available upon reasonable request.