Myeloid cell dynamics correlate with clinical outcomes of severe coronavirus disease 2019

An expanded myeloid cell compartment is a hallmark of severe coronavirus disease 2019 (COVID-19); however, it remains unclear whether myeloid cells are benecial or detrimental to the clinical outcome. Here, we tracked cellular dynamics of myeloid-derived suppressor cell (MDSC) subsets and examined whether any of them correlate with disease severity and prognosis by ow cytometric analysis of blood samples from COVID-19 patients. We observed that polymorphonuclear (PMN)-MDSCs, rather than other MDSC subsets, transiently expanded in severe cases but not in mild or moderate cases. Notably, this subset was selectively expanded in survivors of severe cases and diminished during recovery. Analysis of plasma cytokines/chemokines revealed that interleukin-8 increased prior to PMN-MDSC expansion in survivors and returned to basal levels during the recovery phase. In contrast, interleukin-6 and interferon-γ-induced protein 10 were abundantly induced in non-survivors, suggesting possible downstream targets for the immunosuppressive effects of the MDSC subset. Our data indicate that increased cellularity of PMN-MDSCs might be benecial for the clinical outcome and could be useful as a possible predictor of prognosis in cases of severe COVID-19. test. Average time points for cellular analysis were compared between survivors and non-survivors of severe COVID-19 at the late phase by a two-tailed Student’s t-test, and a p-value < 0.05 was considered signicant. For plasma cytokine levels below the detection limit, the value was set to 0.01 pg ml -1 . Cytokine concentrations were compared with one-way ANOVA with post-hoc Tukey’s honest signicant difference test. Statistical signicance was set at *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Spearman correlations between plasma cytokine concentrations and cell frequencies were identied in all specimens. Correlations with r > 0.4 or r < −0.4 and p < 0.01 were considered signicant. A simple regression line was shown for a signicant correlation. GraphPad Prism version (GraphPad was used for all statistical analyses and graphical representations.


Introduction
In December 2019, coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was rst reported in Wuhan, Hubei Province, China, and rapidly spread causing a global pandemic [1][2][3][4][5] . Although most COVID-19 patients exhibit asymptomatic or mild clinical symptoms that resemble seasonal coronavirus diseases, 19% of patients suffer from severe or critical disease with 2.3% mortality 6 . The countermeasure for the pandemic is prioritized to protect those who are at greater risk of death from COVID-19, including the elderly and those with comorbidities, such as hypertension, diabetes, and cardiac and pulmonary diseases 6,7 .
Myeloid-derived suppressor cells (MDSCs) are a heterogeneous population of immature myeloid cells that are generated during a large array of pathogenic conditions from cancer to obesity and mediate immune suppression 8 . In humans, MDSCs consist of at least three groups of cells, monocytic MDSCs (M-MDSCs), polymorphonuclear MDSCs (PMN-MDSCs), and early stage MDSCs (e-MDSCs), which have been found in peripheral blood mononuclear cells (PBMCs), in addition to bone marrow and in ammatory tissues 8 . MDSCs have been reported in various infectious diseases, including bacterial, fungal, parasitic, and viral infections 9 ; however, their roles in disease pathogenesis are still unclear. Notably, individuals with the aforementioned risk factors for COVID-19 are prone to sustained increased frequencies of MDSCs 8,10 . In addition, elevated levels of interleukin (IL)-6 and IL-8, well-known inducers of MDSCs 11-13 , are observed in severe cases of COVID-19 14,15 . This information suggests a possible link between MDSCs and COVID-19. The expansion of MDSCs and MDSC-like cells has been repeatedly observed in severe COVID-19 patients by several research groups from different countries [16][17][18][19][20][21][22] . However, to gain insight into the roles of MDSCs in COVID-19, it is important to analyze the cell subsets in association with different clinical outcomes (e.g. survivors and nonsurvivors). In this study, we describe the transient but prominent expansion of the PMN-MDSC subset in survivors of severe COVID-19. Along with the cytokine/chemokine levels in plasma, our data suggest the bene cial role of PMN-MDSCs, which potentially suppress excessive in ammation during recovery from severe COVID-19.
The frequencies of each MDSC subset among live PBMCs were tracked with time after symptom onset ( Fig. 1). The results are plotted separately for the early and late phases ( Fig. 2a and 2b). During the early phase, a signi cant reduction in e-MDSC frequencies was observed in moderate II and severe surviving cases relative to those in healthy controls (p = 0.0317, healthy vs. moderate II; p = 0.0175, healthy vs. severe surviving cases), partly similar to the results described in a recent study 22 . Therefore, the loss of e-MDSCs in peripheral blood correlates with the onset of moderate II and severe surviving cases.
Although M-MDSC and PMN-MDSC subsets remained unchanged in the early phase, the frequencies of PMN-MDSCs, rather than M-MDSCs, dramatically increased in the late phase of severe COVID-19 ( Fig. 1b, 1c, 2a, and 2b), reproducing ndings in recent publications [17][18][19][20][21][22] . Further extending previous ndings, we found that the increase in PMN-MDSCs was transient, and the frequencies of PMN-MDSCs returned to basal levels during the recovery phase before discharge (Fig. 1c). Together, we conclude that the PMN-MDSC subset emerges in the peripheral blood stream in correlation with disease severity and declines by the time of discharge.
Subdivision of severe cases into survivors and non-survivors gives us important indications as to how the increased MDSC subset contributes to the clinical outcome and whether this subset is bene cial or detrimental to the prognosis. Notably, a selective increase of the PMN-MDSC subset, rather than the M-MDSC subset, in survivors of severe COVID-19 was observed, whereas the PMN-MDSC subset remained unchanged in non-survivors even in the late phase of severe COVID-19 ( Fig. 1b, 1c, 2a, and 2b). The average time points for cellular analysis in both groups were not signi cantly different (survivor: day 19 ± 5, non-survivors: day 16 ± 4, mean ± SD, p = 0.35, two-tailed Student's t-test). Therefore, defective expansion of the PMN-MDSC subset in non-survivors is unlikely, owing to the analysis time differences.
As IL-6 and IL-8 are key regulators of M-MDSCs and PMN-MDSCs, respectively 11-13 , we rst focused on the correlation between these cytokines/chemokines and the cellularity of MDSC subsets. In the study cohorts, a signi cant change was observed in plasma IL-6 in severe fatal cases (p = 0.0348, healthy vs.  Fig. 3a and 3b) 14,15 . Notably, IL-8 elevation in the early phase was a speci c event in survivors; non-survivors failed to produce as much IL-8 as the survivors (Fig. 3a).
The correlation between IL-6/IL-8 levels and each MDSC subset was subsequently examined. Plasma IL-6 levels were negatively correlated with the e-MDSC subset in this cohort (Fig. 4a), suggesting the possible involvement of IL-6 in the reduction of the e-MDSC subset. Additionally, plasma IL-6 levels were positively correlated with the M-MDSC subset (Fig. 4a). These results are consistent with the previous idea that IL-6 is a key inducer of the M-MDSC subset 11,12 .
We observed a speci c increase in IL-8 levels in the early phase and PMN-MDSC subset in the late phase from survivors of severe cases. These results imply a link between IL-8 and PMN-MDSC induction. In agreement with this, we observed a positive correlation between IL-8 levels and the frequencies of PMN-MDSCs, whereas this correlation was not found for IL-6 ( Fig. 4a and 4b). Given the chemoattractant activity of IL-8 in the PMN-MDSC subset 13 , PMN-MDSCs might be recruited into peripheral blood by IL-8 following the onset of severe COVID-19.
Association between MDSC subsets and plasma IP-10 levels Consistent with recent reports 25,26 , plasma levels of the interferon-γ-inducible chemokine IP-10 were elevated in the early phase of severe fatal cases (Fig. 3a). Similar to IL-6, this chemokine was elevated in non-survivors of severe cases (Fig. 3a). The negative correlation between IP-10 and the e-MDSC subset suggests possible involvement of this chemokine in e-MDSC reduction, in addition to IL-6 ( Fig. 4a and  4c). Since IL-6 and IP-10 were abundantly induced in non-survivors ( Fig. 3a and 3b), their related cells might be possible targets for the immunosuppressive effects of the PMN-MDSC subset.

Discussion
MDSCs are heterogeneous myeloid cell subsets that expand proportionally to the severity of in ammatory diseases. They are equipped with several immunosuppressive molecules; however, whether they are bene cial or detrimental to the clinical outcomes of cancer, obesity, and chronic infectious diseases remains unclear 8,9 . In agreement with this, several groups have recently reported the expansion of MDSC subsets in severe COVID-19 patients [16][17][18][19][20][21][22] ; however, the link between MDSC subsets and clinical outcome remains to be addressed, owing to limitations in experimental design. By separately analyzing survivors and non-survivors of severe COVID-19, we revealed the expansion of the PMN-MDSC subset as a survival-speci c event that is associated with recovery from the disease. Thus, our data favor a bene cial role of the PMN-MDSC subset, contributing to recovery from severe COVID-19.
Upon COVID-19 onset, the levels of the e-MDSC subset declined more substantially in moderate II and severe surviving cases than in the healthy individuals, partly consistent with results of a recent study 22 . Analysis of plasma cytokine/chemokine levels also revealed a negative correlation between e-MDSC frequencies and IL-6/IP-10 levels. The mechanism by which these factors regulate the cellularity of the e-  21 . Thus, the balance between PMN-MDSC and M-MDSC subsets might be crucial for regulating disease severity and could be associated with prognosis. Furthermore, we previously identi ed PMN-MDSC-like cells, known as IFN-γ-producing immature myeloid cells, which are key cellular components that confer protection against severe bacterial infection [28][29][30] . Thus, it is intriguing to speculate that additional PMN-MDSC-like cells might exist that promote recovery from severe COVID-19.
MDSCs are generally considered to play a harmful role in cancer and several infectious diseases 8,9 , whereas they have a bene cial role not only in generating tolerance in autoimmune diseases and allograft transplantation but also in providing protection against sepsis [31][32][33] . During severe COVID-19, neutralizing antibody responses arise slower than they do against in uenza viruses, become detectable after a >1-week incubation period, and peak at 3-4 weeks following symptom onset 34 . Therefore, the period prior to the induction of acquired immunity might be a key decisive point for the clinical outcome of severe cases in which the control of viral replication without hyperin ammation is needed. Along the same line, it has been shown that the anti-in ammatory drugs dexamethasone and anti-IL-6 receptor monoclonal antibody reduce the mortality rate of severe COVID-19 cases [35][36][37][38] . We speculate that the emergence of the PMN-MDSC subset prolongs survival by suppressing excessive in ammation until acquired immunity is introduced. In agreement with the hypothesis that MDSCs can inhibit/reduce the severe lung in ammation/sepsis associated with COVID-19 39 , we propose that the expansion of the PMN-MDSC subset might be bene cial for the clinical outcome and could be useful as a possible predictor of prognosis in cases of severe COVID-19.

Ethics approval and consent to participate
This study protocol was approved by the National Institute of Infectious Diseases Ethic Review Board for Human Subjects (Permit number: 1107 and 1111). All participants provided written informed consent in accordance with the Declaration of Helsinki.

Human subjects
This study enrolled 40 patients with mild (n = 12), moderate I (n = 7), moderate II (n = 8), and severe (n = 13) COVID-19 at two hospitals in Japan (Yokohama Municipal Citizen's Hospital and Center Hospital of the National Center for Global Health and Medicine), as well as seven healthy donors. The severity of symptoms was strati ed according to the third edition of the medical guidelines of COVID-19 provided by the Japanese Ministry of Health, Labor and Welfare 40 . Mild cases were de ned as having no pneumonia and no, or limited, clinical symptoms. Moderate I and moderate II cases were de ned as the presence of pneumonia without and with the need for supplemental oxygen (93% < SpO 2 < 96%, moderate I; SpO 2 ≤ 93%, moderate II), respectively. Severe cases were de ned as having pneumonia and respiratory distress requiring ICU admission or ventilator use. Clinical characteristics of patients and time points of blood sampling are provided in Supplementary Table S1.

Preparation of PBMCs and plasma
Blood samples were collected from the COVID-19 patients and healthy donors using a BD Vacutainer® CPTTM Tube (BD Biosciences, Franklin Lakes, NJ) and centrifuged for 20 min at 1500-1800 g at 23 °C. Following thorough washing, specimens were analyzed using a FACSAria III ow cytometer (BD Biosciences). Data obtained from ow cytometry were analyzed using FlowJo v10.6.1 (BD Biosciences).
Cytokine/chemokinequanti cation Plasma cytokines/chemokines were measured using a cytometric bead array kit (BD Biosciences) according to the manufacturer's instructions. Data were acquired using a FACSCalibur ow cytometer (BD Biosciences) and analyzed using FCAP Array Software Version 3.0 (BD Biosciences).

Statistical analysis
Data comprising the ow cytometric frequencies of cells were compared with two-way ANOVA with posthoc Tukey's honest signi cant difference test. Average time points for cellular analysis were compared between survivors and non-survivors of severe COVID-19 at the late phase by a two-tailed Student's t-test, and a p-value < 0.05 was considered signi cant. For plasma cytokine levels below the detection limit, the value was set to 0.01 pg ml -1 . Cytokine concentrations were compared with one-way ANOVA with posthoc Tukey's honest signi cant difference test. Statistical signi cance was set at *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Spearman correlations between plasma cytokine concentrations and cell frequencies were identi ed in all specimens. Correlations with r > 0.4 or r < −0.4 and p < 0.01 were considered signi cant. A simple regression line was shown for a signi cant correlation. GraphPad Prism version 8.0 (GraphPad Software, San Diego, CA, USA) was used for all statistical analyses and graphical representations.

Page 17/17
This is a list of supplementary les associated with this preprint. Click to download.