Alterations of blood monocyte subset distribution and surface phenotype are linked to infection severity in COVID‐19 inpatients

Abstract Severe coronavirus disease 19 (COVID‐19) manifests with systemic immediate proinflammatory innate immune activation and altered iron turnover. Iron homeostasis, differentiation, and function of myeloid leukocytes are interconnected. Therefore, we characterized the cellularity, surface marker expression, and iron transporter phenotype of neutrophils and monocyte subsets in COVID‐19 patients within 72 h from hospital admission, and analyzed how these parameters relate to infection severity. Between March and November 2020, blood leukocyte samples from hospitalized COVID‐19 patients (n = 48) and healthy individuals (n = 7) were analyzed by flow cytometry enabling comparative analysis of 40 features. Inflammation‐driven neutrophil expansion, depletion of CD16+ nonclassical monocytes, and changes in surface expression of neutrophil and monocyte CD64 and CD86 were associated with COVID‐19 severity. By unsupervised self‐organizing map clustering, four patterns of innate myeloid response were identified and linked to varying levels of systemic inflammation, altered cellular iron trafficking and the severity of disease. These alterations of the myeloid leukocyte compartment during acute COVID‐19 may be hallmarks of inefficient viral control and immune hyperactivation and may help at risk prediction and treatment optimization.


Introduction
Since late 2019, the SARS-CoV-2 pandemic has resulted in a global health crisis. As of March 2022, more than 472 million infections and over 6.1 million COVID-19 deaths were documented [1]. COVID-19 is characterized by a heterogeneous impaired immune control of viral replication leads to a "cytokine release syndrome" with subsequent organ damage [5,6]. Specifically, low levels of CD4 + and CD8 + cells, impaired antiviral type I IFN response and polymorphisms in the genes involved in antigen presentation and immune cell communication were associated with the risk of severe COVID-19 [7][8][9][10]. Furthermore, aberrant monocyte and neutrophil accumulation along with systemic inflammation and altered systemic iron homeostasis have been linked to the disease severity and a poor prognosis [11][12][13][14][15][16][17][18][19][20]. This pertains to the key role of myeloid leukocytes, that is, macrophages, monocytes, and neutrophils in the SARS-CoV-2 pathogen control and orchestration of the inflammatory response [13,16,17,[21][22][23][24]. In addition, monocytes or macrophages play an important role in the orchestration of iron homeostasis, and inflammatory processes result in macrophage or monocyte iron retention, whereas iron exerts distinct effects on macrophage polarization and immune effector functions [25][26][27].
Human blood monocytes consist of three populations: the major subset of CD14 +/high CD16 −/low classical cells, and substantially less abundant CD14 +/high CD16 +/high intermediate, and CD14 −/low CD16 +/high nonclassical monocytes [28]. These subsets display differences in the pattern to activate stimuli and immune effector functions such as phagocytosis, antigen presentation, and cytokine production [28]. Alterations in the monocyte subset distribution and surface molecule repertoire were observed in multiple inflammatory conditions, including COVID-19, and were proposed to correlate with the disease course and prognosis [11, 15-18, 21-23, 28-33].
From the clinical and immunological point of view, there is an ongoing search for markers characterizing the clinical course of COVID-19 as well as the protective and pathological immune responses to the pathogen. Herein, we investigated the surface molecule repertoire of myeloid innate immune cells from hospitalized COVID-19 patients by flow cytometry and linked the surface phenotype to the disease severity and patterns of systemic inflammation and iron turnover by self-organizing map (SOM) clustering [34,35].

Study cohort
The study was conducted between March and November 2020 and encompassed healthy controls (n = 7) and hospitalized COVID-19 patients (n = 53 enrolled) whose blood myeloid compartment was analyzed by flow cytometry within 72 h from hospital admission (Supporting information Figures S1 and S2). Due to insufficient flow cytometry sample quality, five COVID-19 patients were excluded from further analysis. Finally, samples from seven healthy controls, 16 moderate COVID-19 patients not requiring supplemental oxygen therapy during hospitalization (WHO score 3), and 32 severe COVID-19 individuals requiring low/high flow oxygen treatment or mechanical ventilation (WHO score 4-6) were available for the anal- ysis ( Figure 1). Among severe COVID-19 cases, 22% were treated at an intensive care unit and 6.2% deceased during hospitalization. Males constituted 56% of the moderate and 72% of the severe COVID-19 group, but the sex distribution difference was not significant.  Table S1).

Sustained systemic and cellular inflammatory response and iron restriction in COVID-19
Systemic inflammation, as evidenced by elevated levels of IL6, C-reactive protein (CRP), and neopterin, was present in hospitalized COVID-19 individuals and was linked to poor outcomes [12,[14][15][16]. In our study cohort, plasma IL6, CRP, and neopterin concentrations at admission were significantly higher in patients developing severe COVID-19 than in subjects with moderate disease, as described previously [12] (Supporting information Figure S3A). These differences in inflammatory markers were paralleled by significantly reduced circulating iron concentrations, transferrin saturation (TF-Sat), and elevated circulating ferritin levels in severe COVID-19 as compared with the moderate group (Supporting information Figure S3B). Such a phenomenon reflects iron restriction in response to a severe infection, as reported previously for COVID-19 [16,20,[36][37][38][39].

Alterations of the myeloid compartment composition and surface phenotype in healthy volunteers and moderate and severe COVID-19 patients
Next, we investigated how this systemic inflammatory and ironrestricted milieu shapes the composition, inflammatory, and iron transporter protein repertoire of myeloid leukocytes [26,40]. To this end, we analyzed the whole blood monocyte and neutrophil compartments of healthy controls and COVID-19 patients within 72 h of hospital admission with multicolor flow cytometry (Supporting information Table S2). Neutrophils were defined within the lymphocyte lineage-negative leukocyte population (Lin − : CD3 − CD19 − CD56 − ) by logical gating (AND) of the CD16 + , CD11b + , CD62L + , and SSC hi cells. Monocytes were identified within the non-neutrophil gate by UMAP clustering (uniform manifold approximation and projection) [41] in respect to CD11b, CD14, CD15, CD16, CD62L, CCR2, CX3CR1, and HLA-DR expression levels (Supporting information Figure S1). Classical, intermediate, and nonclassical monocyte subsets were defined by differences in CD14, CD16, HLA-DR, CCR2, and CX3CR1 surface levels [28] within the monocyte cluster cells (Supporting information Figure S2). Of note, the additional markers used to define monocyte populations (HLA-DR, CCR2, and CX3CR1) displayed similar expression differences between classical, intermediate, and nonclassical cells from healthy donors and COVID-19 patients. No downregulation of HLA-DR levels in classical monocytes, described previously [11,16,17], could be observed in COVID-19 patients as compared with healthy controls (Supporting information Figure S4). In turn, neutrophil CD14 expression was found significantly different between the study groups and was the highest in moderate COVID-19 individuals (Supporting information Figure S5).
The neutrophil CD64 was virtually absent in healthy controls ( Figure S8).

Patterns of innate myloid leukocyte response in acute COVID-19
To elucidate possible patterns of coregulation of myeloid leukocyte quantity and surface phenotype, we subjected the set of 40 flow cytometry parameters measured in our study (Supporting information Table S3) and the healthy and COVID-19 study participants to SOM clustering [34,35].
Healthy controls and moderate COVID-19 cases comprised 75% of the largest participant cluster #1 which was characterized by the lowest levels of CD40/CD64/CD71 parameters as well as healthy-like monocyte subset distribution, neutrophil levels, and neutrophil CD64 expression ( Figure 4A,B). Interestingly, none of the severe COVID-19 cases assigned to the cluster #1 required Cytometry feature regulation n = 51 -55 ICU treatment (Supporting information Table S4). The remaining clusters consisted predominantly of severe COVID-19 patients with increasing frequency. The hallmark of the cluster #2 was an increased expression of CD86, C163, the immunosuppressive PD-L1 molecule (CD279), and the iron exporter FPN1 in classical and intermediate monocytes as well as an expansion of the intermediate subset. The participant cluster #3 in turn was primarily characterized by an elevated neutrophil and monocyte CD64 as well as a depletion of the nonclassical monocytes. Finally, leukocyte samples assigned to the cluster #4 displayed a consistent expansion of the neutrophil compartment, elevated CD64 in neutrophils, and a depletion nonclassical monocytes ( Figure 4A). In comparative analysis of COVID-19 individuals, the participant clusters demonstrated similar age and body mass index. In turn, the clusters #1 and #4 COVID-19 individuals were predominantly males (78%) as compared with the remaining subjects. Concerning the disease course, the fraction of participants requiring respiratory support was the lowest in cluster #1 and peaked in the clusters #3 and #4. However, none of these clinical features was significantly different in the all-group and the cluster #1 versus rest comparisons (Supporting information Table S4).
Finally, we investigated the inflammatory status and iron turnover parameters in COVID-19 patients assigned to the participant clusters. Although the differences between the clusters did not reach statistical significance following multiple testing correction, a clear tendency toward increased blood levels of IL6, CRP, and neopterin was observed in the clusters #2, #3, and #4 as compared with the healthy-like cluster #1 ( Figure 5, Supporting information Table S4). Accordingly, the clusters #2, #3, and #4 had the lowest levels of circulating iron and TF-Sat values suggestive of inflammatory iron restriction [36]. Of interest, despite the elevated inflammatory marker levels, blood ferritin concentration in the cluster #2 tended to be lower than in the similarly inflammatory clusters #3 and #4 ( Figure 5, Supporting information Table S4).
Collectively, the clustering analysis results suggest the presence of qualitatively and quantitatively distinct phenotypes of innate myeloid cell response to SARS-CoV-2 infection. Such phenotypic diversity is likely associated with COVID-19 severity and partially independent of the demographic patient background.

Discussion
Exaggerated cytokine and innate cell response is hallmark of severe COVID-19, requiring hospitalization, oxygen, or ICU   Table S3) and study participants (n numbers presented in B) were subjected to clustering analysis with a combined self-organizing map (SOM)-hierarchical clustering (HCl) algorithm (Supporting information Figure S9). Three clusters of cytometry features termed after the most characteristic features (neutrophil, CD40, CD64, CD71, CD86, and CD163) and four participant clusters were identified. (A) Normalized levels of flow cytometry parameters (Z-score) in the participant (horizontal) and cytometry feature clusters (vertical plot facets) presented as a heat map. N = 48 biological replicates (blood cell donors, cluster #1: n = 16, #2: n = 12, #3: n = 11, #4: n = 9). (B) Percentages of healthy controls, moderate and severe COVID-19 participants in the participant clusters. Statistical significance was determined by χ 2 test. N = 48 biological replicates (blood cell donors, cluster #1: n = 16, #2: n = 12, #3: n = 11, #4: n = 9). therapy [11,12,[14][15][16][17][18][19]. In agreement with previous reports, we observed elevated circulating IL6, CRP, and neopterin levels and an expansion of neutrophils with a concomitant reduction in lymphocytes in COVID-19 patients requiring respiratory support. Similar to other infections [36,38], this resulted in reduced levels of circulating iron and elevated ferritin, in line with few published reports on COVID-19 [16,20,27,37,37,39]. The sole cellular iron exporter, FPN1, in monocytes and macrophages, is downregulated by the hormone hepcidin and diverse inflammatory cytokines and upregulated by cellular iron [25,36,42,43]. Interestingly, hypoferremia in acute COVID-19 was accompanied by only mildly elevated blood hepcidin levels [37,39], which may explain the consistently elevated FPN1 levels in the subsets of inflammatory COVID-19 patients assigned to the participant cluster #2. In general, the changes in the cellular iron trafficking and systemic turnover in COVID-19 may bear relevance for the effective SARS-CoV-2 control because iron levels significantly impact on the differentiation of T cells and monocytes and on macrophage polarization and also affect antimicrobial immune effector mechanisms [25,36]. Furthermore, intracellular iron availability may promote viral replication [44]. As such, the interplay of these processes requires further mechanistic and clinical investigations.
By applying flow cytometry, we were able to characterize a palette of 40 variables associated with the quantity, pro-, antiinflammatory, and iron transport phenotype of neutrophils, classical, intermediate, and nonclassical monocytes. Among them, a consistent depletion of nonclassical monocytes, in part paralleled by an expansion of the intermediate subset became evident in COVID-19 patients. This phenomenon was also described by others and suggested to aggravate with infection severity [11, 15-18, 21-23, 31-33]. It was also proposed to predict the risk of critical disease and respiratory failure [17,21,23]. The CD16 + nonclassical subset contraction was suggested to be specific for COVID-19 but not for bacterial sepsis [16]. However, results of an experimental endotoxin challenge in humans demonstrated a general depletion of blood monocytes and a step-wise reconstitution involving immediately the classical population, followed by the remaining subsets [30]. Hence, the nonclassical monocyte contraction in COVID-19 may reflect an early step in the nonspecific inflammatory response to SARS-CoV-2. A subsequent repopulation of the nonclassical compartment by classical monocytederived precursors during convalescence can be inferred from longitudinal COVID-19 immunophenotyping studies [16][17][18].
Our immune-phenotyping results revealed an upregulation of the FC-γ receptor CD64 in neutrophils, classical and intermediate monocytes in COVID-19 patients as described recently [11,17,19,32,45]. For neutrophils, the elevated CD64 expression together with lowered neutrophil CD14 levels observed in the study group of severe COVID-19 patients may indicate the presence of less mature proneutrophil-like CD64 + CD14 low cells expanding in early COVID-19 [11] and sepsis [46,47], and hence, represent a general reaction to a strong inflammatory stimuli triggering emergency myelopoiesis [11,17]. This is further supported by our clustering data, where the high neutrophil CD64 levels co-occurred with the neutrophil expansion in the highly inflammatory, severe COVID-19-dominated participant cluster #4. Since the CD64 expression in neutrophils and monocytes was the highest in the clusters #3 and #4 comprising of predominantly ventilated severe COVID-19 patients, secondary bacterial airway coinfections may also contribute to the upregulation of this marker. The expression of CD86 in the intermediate and classical monocyte subsets followed a distinct regulatory pattern with a peak in moderate COVID-19. which is in line with other publications [14,22]. In the clustering analysis of our cytometry data, monocyte CD86 expression was found to be coregulated with the immunosuppressive ligand CD279 (PD-L1), the alternative macrophage activation marker CD163 and the iron exporter FPN1, the latter being highest in the participant cluster #2. Collectively, this suggests that in a fraction of COVID-19 patients, subsumed under the participant cluster #2 in the study cohort, the blood monocytes may acquire an immunoregulatory, T-cell suppressive, iron export CD86 high CD279 high FPN1 high phenotype. Notably, expansion of a similar HLA-DR low , CD86 hi , CD163 hi , and T cell-suppressive classical monocyte populations was described for moderate and severe COVID-19 [11,[16][17][18]24]. In our cohort, however, no significant differences in monocyte HLA-DR expression between COVID-19 patients and healthy controls could be observed, which may be attributed to the differences in the detection methodology (multicolor flow cytometry versus CyTOF or spectral flow cytometry).
To get a thorough overview of myeloid leukocyte response in acute COVID-19, we performed a clustering analysis with the SOM technique [34,35], by which we could identify four distinct clusters among the study participants. The largest, healthy-like cluster #1 was associated with near-normal composition and surface phenotype of neutrophils and monocytes as well as the lowest levels of systemic inflammation, largely normal iron homeostasis and low risk of ICU admission. The other clusters were characterized by a clearly inflammatory and iron-restrictive phe-notype but differed in the myeloid leukocyte characteristic. In particular, the cluster #2 was hallmarked by the presence of CD86-, CD163-, CD279-, and FPN1-expressing monocytes suggestive of an immunoregulatory phenotype [11,[16][17][18]24] as discussed above. The cluster #3 was characterized primarily by high levels of monocyte CD40 and CD64 which may reflect a proinflammatory priming of those cells, as observed in chronic inflammatory conditions [48][49][50]. Finally, the expansion of CD64expressing neutrophils in the participant cluster #4 may represent an inflammation-triggered emergency myelopoiesis and occurrence of immature neutrophils in the circulation of severe COVID-19 patients [11,17,46,47]. The prognostic and therapeutic relevance of such individual heterogeneity of the innate response to the SARS-CoV-2 pathogen needs to be investigated in a larger study collective.
Our study bears limitations referring to a small number of participants which, together with the lack of verification cohort, precluded more systematic analyses adjusted for known risk factors of severe COVID-19 such as age, sex, and comorbidity. The alterations in neutrophil and monocyte subset quantities and phenotype as well as the clinical utility of the predictive myeloid score needs to be verified in a larger follow-up study, possibly including longitudinal determinations to capture dynamics of the myeloid leukocyte compartments [15][16][17][18]22]. Additionally, differences in age structure and likely in the frequency of comorbidity between the healthy controls and COVID-19 patients in the study cohort may have confounded the presented analysis results.
Our study corroborates the massive changes in the neutrophil and monocyte compartments in acute moderate and severe COVID-19 including expansion of neutrophils, depletion of the nonclassical monocytes, as well as regulation of CD64 and CD86 expression and altered iron trafficking in myeloid leukocytes. Finally, we demonstrate the presence of four patterns of the innate myeloid response to SARS-CoV-2 with possible implications for diagnosis, risk assessment, and therapy of COVID-19.

Study population and procedures
The study was performed between March 21, 2020 and November 16, 2020 at the hospital of the University of Innsbruck, Tyrol, Austria. The inclusion criteria for COVID-19 participants were age ≥18 years and hospital admission because of acute SARS-CoV-2 infection confirmed by nasal or oral swab PCR test. The exclusion criterion was a critical physical or mental state at admission, precluding giving informed study participation consent. Healthy controls were recruited among the institution's health care workers.
Measurement of the laboratory inflammation markers: CRP, IL6, neopterin, and iron status parameters: ferritin, iron concentration, and TF-Sat, in the blood at admission was done as a routine determination at the hospital central laboratory. Whole venous EDTA blood (5-10 mL) for the flow cytometry measurement was withdrawn within 72 h from hospital admission.

Measurement of laboratory inflammation and iron markers
CRP in plasma was determined by a particle-enhanced immunological turbidity assay: human CRP agglutinated with latex particles coated with anti-CRP mAb (mouse). Aggregates were determined turbidimetrically. IL-6 was assessed in plasma by an electrochemiluminescence assay in the sandwich principle. Neopterin was analyzed by a competitive enzyme immunoassay for the direct quantitative determination of neopterin. Quantitative determination of ferritin in human plasma was performed using a latex particle-enhanced immunological turbidity assay. The amount of iron in human plasma was measured photometrically by the FerroZine method without deproteinization. TF-Sat was measured by an immunological turbidity test.
Whole blood was subjected to RBC depletion with ammoniumchloride-potassium buffer, followed by extensive washing with FACS buffer (1% FCS, 0.5 mM EDTA in PBS) and staining with the backbone antibody mix (dilutions: 1:20-1:200 in BD Horizon Brilliant Stain Buffer, Beckton-Dickinson) for 10 min at 4°C. Next, the sample was washed, split, and stained with the separate isotype or test antibody mix (1:200 dilution in BD Horizon Brilliant Stain Buffer) for an additional 10 min at 4°C. The samples were measured with a Cytoflex S device (Beckman Coulter, Brea, CA) and analyzed with FlowJo 10.7 software (Beckton-Dickinson).
The cytometry data were analyzed by an experienced operator blinded to the study group assignment with a strategy depicted in Supporting information Figures S1 and S2 according to flow cytometry guidelines [52]. Neutrophils were identified within the lineage-negative leukocytes (CD3 − CD19 − CD56 − ) by logical gating (AND) of CD11b + , CD16 + , CD62L + , and SSC lo events. Monocytes were defined within the non-neutrophil gate (logical NOT gating) by UMAP [41] in respect to CD11b, CD14, C15, CD16, CD62L, CCR2, CX3CR1, and HLA-DR expression. The monocyte cluster was identified by high CD14, CCR2, CX3CR1, and HLA-DR levels. Classical, intermediate, and nonclassical monocytes were identified within the monocyte UMAP cluster by differences in CD14, CD16, CCR2, CX3CR1, CD62L, and HLA-DR expression [28]. Surface expression was expressed as a difference in median signal intensity ( MFI) between the specific staining and the isotype. Neutrophil CD71 and CD163 as well as CD80 staining in neutrophils and monocytes were excluded from the downstream analysis due to low and virtually invariant expression levels. MLR was calculated as a ratio of the CD45 + percent of HLA-DR + panmonocytes to Lineage + (CD3, CD19, CD56) cells. NLR was calculated as a ratio of the CD45 + percent of CD15 + SSC hi neutrophils to Lineage + cells.
Since most of the analyzed variables demonstrated nonnormal distribution (Shapiro-Wilk test), nonparametric hypothesis tests were utilized in the downstream analyses. To search for cytometry features (Supporting informationn Table S3) significantly differing between healthy controls, moderate and severe COVID-19 patients, Kruskal-Wallis test with with Benjamini-Hochberg (FDR: False Discovery Rate) correction for multiple comparisons was applied [55]. The significant variables were defined by the pFDR < 0.05 cutoff. Post-hoc testing for pairwise differences between the groups was accomplished by Benjamini-Hochberg corrected Mann-Whitney U test. Clustering of the flow cytometry features (Supporting information Table S3) and of the healthy and COVID-19 study participants in respect to the flow cytometry features was done with a combined SOM and hierarchical clustering (HCl) procedure [34,35]. The data were preprocessed by Z-score normalization and the participants with missing values eliminated. The cytometry features or participants were subjected to dimensionality reduction by SOM (5 × 5 hexagonal grid, cosine distance between the features/participants) and, subsequently, the SOM nodes were clustered with the HCl (cosine distance between the nodes, Ward D2 algorithm). The choice of the SOM node cluster number was motivated by the bend of the within-cluster sum-of-square curve and a visual analysis of the dendrograms (Supporting information Figure S9). The fraction of clustering variance (ratio of the total within-cluster to total sum of squares) was 0.38 and 0.58 for the feature and participant clustering, respectively. Support by the Christian Doppler Society (GW), the Austrian Research Fund Doctoral Programm HOROS (FWF-W1253, GW, VP), the intramural funding program of the Medical University Innsbruck for young scientists MUI-START, Project 2021-01-001 (DH), and the Society for Funding Research and Education in Molecular Immunology and Infectiology ("Verein zur Förderung von Forschung und Weiterbildung in Molekularer Immunologie und Infektiologie") is gratefully acknowledged.

Ethics approval:
The study was performed in accordance with the Declaration of Helsinki and the European Data Policy and approved by the institutional review board of the University of Innsbruck (approval number: 1091/2020).
Patient consent: All study participants were conscious and gave a written informed consent prior to enrollment. The participants' data were anonymized after recording hence making identification of the particular subjects impossible.

Conflict of interest:
PT now owns his data science enterprise, Data Analytics as a Service Tirol. The author declares no other conflict of interest. Other authors declare that non conflict of interest exists.

Data availability statement:
The data that supports the findings of this study are available in the supplementary material of this article. The R analysis pipeline is available at https://github.com/PiotrTymoszuk/covFacs. The data set with variables analyzed in the study is available as Supporting information Table S5.