Circulating haematopoietic stem cells and long‐term outcomes of COVID‐19

An acute depletion of circulating haematopoietic stem/progenitor cells (HSPCs) occurs during COVID‐19, especially among patients with a poorer disease course. We herein examined whether HSPCs levels at hospital admission for COVID‐19 predict 1‐year mortality and the long‐COVID syndrome.


| INTRODUCTION
The COVID-19 pandemic was responsible for over 770 million confirmed cases and over 6.9 million deaths worldwide as of November 2023.Hospitalizations, admissions to intensive care unit and deaths due to COVID-19 are still occurring. 1As of today, COVID-19 still represents a major health concern.
Since the beginning of the pandemic, it has been realized that some patients develop sign and symptoms that persist for months after the acute SARS-CoV-2 infection, the so called long-COVID or post-COVID syndrome. 2ong-COVID is a heterogeneous condition affecting potentially every organ. 3Its prevalence varies across studies and according to the criteria used, but is estimated to occur in 10-30% of nonhospitalized cases, and up to 50-70% in hospitalized patients, whereas it is less common in vaccinated individuals. 3Specific diagnostic tests and consensus definition for long-COVID syndrome are lacking.According to National Institute for Health and Care Excellence (NICE) guideline, long-COVID can be defined as signs and symptoms that develop during or after the SARS-CoV-2 infection and continue for more than 12 weeks, not being explained by an alternative diagnoses. 4Signs and symptoms include, among other, fatigue, dyspnoea or shortness of breath, chronic cough, insomnia, anxiety or depression, cognitive impairment and difficulty of concentration.It usually presents with clusters of symptoms, often overlapping, which can change over time and fluctuate in severity. 4Risk factors for the development of long-COVID differ across study.][7] The pathogenesis of this condition is still unknown.Several pathogenic pathways have been hypothesized, including the persistence of virus or its RNA and proteins, autoimmunity triggered by the infection, reactivation of latent viruses, microvascular clotting with endothelial dysfunction and organ/tissue damage trigged by chronic low-grade inflammation. 8The discovery of biomarkers of long-COVID can be helpful to identify patients at risk, monitoring the condition and understanding better its mechanistic drives.Haematopoietic stem/progenitor cells (HSPCs) are a subtype of white blood cell involved in maintaining haematopoiesis, regulating immune surveillance and aiding peripheral tissue homeostasis, including vascular repair and regeneration.HSPCs are mobilized from the bone marrow (BM) after stressor events. 9,10ow levels of HSPCs have been associated with diabetes, the metabolic syndrome, end-stage renal disease and frailty, [11][12][13] and predict the development of chronic complication of diabetes, 14,15 as well cardiovascular events, and death in different populations. 16We previously demonstrated that, in a cohort of patients hospitalized with SARS-CoV-2 infection, HSPCs were dramatically reduced, especially in patients with adverse outcomes (admittance to the intensive care unit or death). 17This led us to hypothesize that pauperization of HSPCs as a mediator of the patient characteristics driving tissue damage in COVID-19.
The aim of this study was to evaluate whether HSPC levels at time of hospital admission for SARS-CoV-2 infection predicted long-term outcomes for COVID-19, including 1-year mortality and the development of the long-COVID syndrome.

| Study participants
The study design and data collection have been described previously in detail. 17Briefly, this prospective observational study was conducted according to the principles of the Declaration of Helsinki and approved by the University Hospital of Padua Ethical Committee (No. 4930/Ao/20).Subjects were consecutively recruited between December 2020 and March 2021 among those admitted to the infectious disease ward.Inclusion criteria were age ≥ 18 and symptomatic polymerase chain reaction-confirmed SARS-CoV-2 infection.Exclusion criteria encompassed age of 100 years or older, severe renal or liver disease, advanced cancer, known haematological disorders potentially affecting HSPC count, short life expectancy, pregnancy or lactation, or inability to provide informed consent.For all subjects, we recorded the following characteristics: age, sex, ongoing medications, laboratory data and comorbidities (diabetes, hypertension, obesity, cardiovascular disease, chronic kidney disease, chronic obstructive pulmonary disease, previous cancer, atrial fibrillation, smoking).
Scoring systems developed to predict mortality in critically ill patients and validated also for COVID-19 were calculated for each patient: the National Early Warning outcome, SARS-CoV-2, stem cells Score (NEWS) and its modification, NEWS-2 and the 4C mortality score produced by the ISARIC 4C consortium. 18,19These models primarily rely on patients' characteristics (age, sex, comorbidities), vital signs and level of consciousness (Table S1).

| Measures and Outcomes
A 5 mL venous blood sample was collected at enrolment and freshly analysed for the quantification of HSPCs.150 μL of peripheral blood was stained with the following monoclonal antibodies: 10 μL of anti-human CD45 (Beckman Coulter), 20 μL of anti-human CD34 (Becton Dickinson Biosciences) and 5 μL of CD133 (Beckman Coulter).At least 1 × 10 6 events were acquired in the mononuclear cell fraction.The same trained operator performed the analysis throughout the study.Different definitions of HSPCs were used (total CD34 + , total CD133 + , CD34 + CD133 + and CD34 + CD45 dim ) and reported as relative (/10 6 blood cells) or absolute (/mL of blood) count.The primary end point of the outcome analysis was mortality at 1 year.The secondary outcome was the occurrence of long-COVID syndrome, defined as persistence of symptoms and signs at least 6 months after COVID-19 diagnosis.Outcomes were ascertained by accessing the patients' electronic medical records and the local death registry.

| Statistical analysis
Sample size was calculated for the achievement of the primary endpoint, the impact of HSPC CD34 + on mortality at 12 months.We planned to divide the population in two groups according to the median value of HSPC CD34 + .We calculated that n = 100 patients were sufficient to detect a risk ratio of 2.5 with alpha = 0.05, beta = 0.20.Continuous variables are expressed as mean and standard deviation.Categorical variables are presented as numbers and percentages.Normality was checked using the Kolmogorov-Smirnov test, and non-normally distributed variables were log-transformed before analysis.Comparisons between two groups were analysed using the two-tailed unpaired Student's t test for continuous variables and the chi-square test for categorical variables.Subjects were stratified according to the median value of HSPCs.In these groups, one-year survival rates were estimated by Kaplan-Meier's method and compared with the log-rank test.The Cox proportional hazards model was used to evaluate the predictors of clinical outcomes at 1-year.Variable significantly associated with 12-month mortality was introduced in a multivariable Cox proportional hazards models.To avoid multicollinearity among HSPC phenotypes, different models were explored, each utilizing a single HSPC phenotype.Results were expressed as hazard ration (HR) and 95% confidence interval (CI).Univariate and multivariate linear regression analyses were run to test the association between the presence of long-COVID symptoms and variables at hospital admission.Results were expressed as coefficient beta and 95% CI.Statistical significance was accepted at p < 0.05, and all tests were 2-tailed.Statistical analysis was performed using SPSS version 24.0.

| Patient characteristics and mortality at 12 months
The clinical characteristics of the study population (n = 100) at hospital admission are summarized in Table 1 and divided by vital status at follow-up.Participants were on average 67.8-year-old, and 59% were male.At 12 months after index hospitalization, 20 patients had died.Patients who died were significantly older at baseline (76.5 vs 65.6; p = 0.013) and had a higher prevalence of pre-existing coronary heart disease (CHD) and chronic kidney disease (CKD) (p = 0.011 and p < 0.001, respectively) compared to those who were alive at 1 year.Fasting plasma glucose (FPG) at hospital admission was significantly higher in those who died during follow-up observation (185.5 vs 128.2 mg/dL; p = 0.018).Patients who died also showed higher peak C-reactive protein (CRP) and more often required high-flow oxygen and/or invasive ventilation during hospital stay, reflecting a more severe illness.As expected, subjects who died during the followup had significantly higher values of NEWS, NEWS-2 and 4C Mortality score at hospital admission than those who were alive at 12 months, confirming the good performance of these scores in predicting mortality in patients with COVID-19 (Figure 1).
Participants were then divided in two groups according to the median value of HSPCs at the time of hospital admission.Those with lower level of HSPCs had a significantly higher value of NEWS, NEWS-2 and 4C Mortality score than those with higher levels of all HSPC phenotypes (Figure 3).Subjects with lower level of HSPCs at enrolment showed a significantly lower probability of survival at 1 year than those with higher levels of HSPCs regardless of their antigenic phenotype (Figure 4).

| Long-COVID and its predictors
During follow-up, among the patients who survived, 36 reported that at least one symptom persisted for at least 6 months after COVID-19 diagnosis and were classified as having the long-COVID syndrome.Tiredness or fatigue (56%), dyspnoea (42%), joint or muscle pain (28%), 'brain fog' (25%), depression or anxiety (22%) and cough (22%) were the most common symptoms.Patients with long-COVID required more radiological examinations and/or pulmonary function tests and more pulmonology or infectious disease specialist visits than those who had fully recovered (83.3 vs 24.5%; p = 0.001 and 83.3 vs 26.5%; p < 0.001 respectively).Patients were divided in two groups according to the presence/absence of long COVID (Table 3).Those with long-COVID syndrome were significantly younger (61.3 vs 68.5; p = 0.049), more often had had COVID-19 related pneumonia during the hospital stay (97 vs 76%; p = 0.006) and required in-hospital treatment with glucocorticoids (97% vs 78%; p = 0.011) or convalescent plasma (56 vs 24%; p = 0.006), reflecting a more severe illness.Prevalence of pre-existing comorbidities, previous medications, laboratory exams, length of hospitalization and proportion of intensive care unit admittance were similar between the two groups.No differences were also found in the value of NEWS, NEWS-2 and 4C Mortality score and in the levels of HSPCs at admission between the two groups.The incidence of long-COVID was similar in the group with low as compared to that with high HSPCs (CD34 +  44 vs 55% p = 0,826; CD34 + CD133 + 44 vs 55% p = 1.000;CD34 + CD45 dim 50 vs 50% p = 1.000).

| DISCUSSION
We herein show that, in a cohort of patients with COVID-19, lower level of HSPCs at the time of hospitalization predicted long-term mortality independently from confounders.However, HSPCs were not significantly associated with the development of signs and symptoms of long-COVID during the 12 months after hospitalization.
To date, clinical outcome data of patients discharged after SARS-CoV-2 infection are scant and most of the studies focused on mid-term sequelae.Long-term survival still remains largely unknown.
We previously demonstrated in the same cohort of patients that HSPCs reduction predicted an increased risk of short-term adverse outcomes (admittance to the intensive care unit or in-hospital death) and mortality at 6 months.Moreover, reduced HSPC levels were mediators of the negative prognostic effect of hyperglycaemia on COVID-19 outcome. 17With the present  T A B L E 3 (Continued) study, we confirm the role of the reduction of HSPC as independent predictor of long-term adverse outcome of COVID-19.
The relationship between hyperglycaemia and COVID-19 is bidirectional.Hyperglycaemia is a significant predictor of adverse outcomes in patients with COVID-19. 20Notably, prior studies have also shown that SARS-CoV-2 infection may affect βcell function and survival and may induce insulin resistance, resulting in glucose homeostasis abnormalities that persist for months after the recovery. 21,22However, in our study, FPG was associated with 1-year mortality at univariate analysis, but when it was entered into multivariate models, the predictive value of FPG was no longer significant.
The mechanisms potentially accounting for circulating HSPC pauperization observed in subjects with SARS-CoV-2 infection include a direct infection of HSPC by the virus.4][25] Severe COVID-19 is also characterized by dysregulation of haematopoiesis in the BM resulting in anaemia, lymphopenia, thrombocytopenia and increased inflammatory myelopoiesis. 26e have demonstrated that an excess of myelopoiesis in mouse models of diabetes is associated with defective HSPC mobilization from the BM into peripheral blood and that there is an inverse correlation between myelopoiesis markers and circulating HSPC levels. 27Finally, the third hypothesis suggests that the dramatic pauperization of HSPC could be the results of the cytokine storm and the excessive immune response. 28,29Whatever the mechanism, it is well-established that HSPC contributes to local inflammation and immunosurveillance and to tissue homeostasis, including vascular repair. 9,10Therefore, it is plausible that a defective tissue homeostasis and repair could translate into adverse long-term outcomes.
Conversely, the factors driving the occurrence of long-COVID remain controversial.Some studies have shown that pre-existing diabetes is a risk factor for the development of signs or symptoms included in the long-COVID definition.On the contrary, at least 14 other studies, in line with our data, have reported no significant association between diabetes and the incidence of long COVID. 30onsistently with the literature we observed that subject with long-COVID were younger and had had a more severe infection, more often had pneumonia and required an in-hospital treatment with glucocorticoids. 2,31,32owever, we did not find an association between sex, obesity, or other comorbidities and the incidence of long COVID. 2,31,32Furthermore, neither ICU admittance nor in-hospital length of stay predicted the occurrence of long-COVID.Patients with long-COVID appear to have immune abnormalities including abnormal T-cell phenotype and function and abnormal cytokine profile that can be reversed with PD-1 blockade. 33We hypothesized that, being closely associated to in-hospital disease severity and long-term outcome, HSPCs could also represent a biomarker of haemato-immune disruption that precedes and predicts long-COVID.Contrary to such expectation, there was no association between circulating HSPCs levels at admission and the subsequent development of long-COVID.Therefore, HSPCs appear to be influenced by the acute phases of infection and their perturbation reflect susceptibility to short-and long-term mortality.On the other side, they are unrelated to the multi-system sequelae and complains that develop after resolution of the acute phase.
Our study has some limitations worth noting.First, the relatively small simple size may have affected the statistical power of some analyses investigating predictors of long-COVID.Second, this study was conducted only in a hospitalized population which may over-represent those with comorbidities.Third, outcomes and diagnostic criteria for long-COVID vary widely the lack of a consensus on the definition could affect the results. 34Conflicting results among the studies investigating the risk factors for long-COVID may arise from these differences.Also, we did not evaluate HSPC maturity, commitment, and function, their differentiation capacity, their pro or anti-inflammatory profile.A better characterization of HSPC population may be relevant to better understand the association between HSPC and negative outcomes.Finally, we do not know how long the HSPC pauperization last over time after resolution of acute infection, which would be important to understand their potential involvement in long-COVID.
Notwithstanding such limitations, strengths of our study may reside in its novelty.Still, no studies have investigated the impact of HSPCs on long-term outcomes of COVID-19.
We acknowledge that the measurement of HSPC may not be currently available in all clinical centres.Nevertheless, identifying biomarkers for adverse outcomes, such as HSPCs, in addition to the standard haematological profile, could assist clinicians in early identification of subjects at a high risk of adverse outcomes.This, in turn, enables proactive planning for the allocation of healthcare resources.
Further studies with a longer follow-up and a larger sample size are needed to confirm our results and find risk factors for long-COVID.
In conclusion, current and previous analyses suggest that HSPCs appear to play a role in both short-and longterm adverse outcomes of COVID-19 and may represent a valid predictor of 1-year mortality.Our data so far do not the hypothesis that reduced HSPC levels represent risk factors for the occurrence of long-COVID.

AUTHOR CONTRIBUTIONS
BMB, MM, PF, JZ, SM, AF and DB collected and analysed data.BMB, AA and GPF wrote the manuscript, RC, PF, JZ, AC, SM, AF, MP and DB revised the manuscript.AC, AA, DB and GPF designed the study.All authors approved the final version of the manuscript.GPF is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

T A B L E 1 F I G U R E 2
(Continued)  (HR 6.22 95% CI 2.51-15.37p < 0.001) and the need of high-flow oxygen and/or invasive ventilation during hospital stay (HR 4.70 95% CI 1.71-12.95;p = 0.003).In addition, age, FPG, peak CRP, alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were significant predictors of mortality.The reduction of 1-SD in the level of circulating HSPCs was associated with a 3-to 5-fold increase in the risk of 1-year mortality, varying according to the different phenotype used.When these variables were entered into a multivariate Cox proportional F I G U R E 1 COVID-19 severity scores and and 1-year mortality.NEWS (A), 4C-mortality score (MS) (B) and NEWS-2 (C) scores in COVID-19 patients grouped according to their status (live/dead) at 12 months after hospitalization.Column heights represent the mean, and bars represent the SE.Superimposed data points are referred to individual patients.p values are shown for the between-group comparisons.Levels of circulating HSPCs at enrolment in patients with COVID-19 and 1-year mortality.(A) Levels of CD34 + HSPCs at enrolment in COVID-19 patients according to their status (live/dead) at 12 months after hospitalization.(B) Levels of CD34 + CD133 + HSPCs at enrolment in COVID-19 patients according to their status (live/dead) at 12 months after hospitalization.(C) Levels of CD34 + CD45 dim HSPCs at enrolment in COVID-19 patients according to their status (live/dead) at 12 months after hospitalization.Column heights represent the mean and bars represent the SE.Superimposed data points are referred to individual patients.P values are shown for the between-group comparisons.F I G U R E 3 COVID-19 severity scores and levels of circulating HSPCs.NEWS (A), 4C-mortality score (MS) (B) and NEWS-2 (C) scores in COVID-19 patients grouped according to the levels of HSPC phenotypes (below/above the median).Column heights represent the mean, and bars represent the SE.Superimposed data points are referred to individual patients.p Values are shown for the between-group comparisons.

F I G U R E 4
Probability of 1-year survival in subjects with COVID-19 according to the levels of circulating HSPCs.Survival curves in COVID-19 patients according to the levels of CD34 + (A), CD34 + CD133 + (B) or CD34 + CD45 dim (C) HSPCs (below/above the median value).p Values are shown for the between-group comparisons with the log-rank test.T A B L E 2 Predictor of 1-years mortality.

T A B L E 3
Clinical characteristics of participants at admission and long-COVID.
Clinical characteristics of study participants at baseline and 1-year mortality.
Note: Patients were divided according to the presence of Long-COVID.Data are presented as mean (standard deviation) or as percentage, where appropriate.*p < 0.05, # p < 0.01.