Blood from 23 healthy controls (HC) was obtained via the “mini donor service” at the University Medical Center Utrecht (UMCU, Utrecht, the Netherlands). This service provides blood from healthy volunteers for research purposes. Healthy controls gave consent for blood withdrawal and the protocol is approved by the Medical Ethical Committee of the UMCU under study approval number 18/774. Healthy controls were in good health as determined by self-rapportage. Controls were from both sexes and between 18 and 65 years old, as described before (17). See also Table 1 . Samples from the trauma and COVID-19 cohorts were compared to these healthy controls.

Blood- and bone marrow samples were obtained from 10 healthy male volunteers between the age of 18–30 years (see Table 1 ), participating in an experimental human endotoxemia study (ABR NL61136.091.17). Six study participant received a first LPS challenge and five of those participants received a second LPS challenge one week later. Material was gathered before LPS challenge (baseline), four hours after the first LPS challenge and four hours after the second LPS challenge. Study approval was obtained by the ethics review board of the Radboud University Medical Center in Nijmegen, the Netherlands. Written informed consent was obtained from all study participants. Subjects underwent a health screening consisting assessment of medical history, physical examination, electrocardiography and hematological laboratory values. Subjects taking prescription drugs were excluded from the study.

The bone marrow sampling was performed by aspiration from the posterior iliac crest under local anesthesia. A total volume of 40 mL was collected per aspiration into syringes prefilled with sodium heparin. Blood samples were drawn from an arterial catheter (radial artery), using sodium heparin as an anticoagulant. The first blood- and bone marrow sample was obtained prior to the first LPS administration (baseline) and the second blood- and bone marrow sample four hours after both the first and the second LPS challenges. The LPS challenges were performed as published previously by Kiers et al. (18). In short, the subjects were infused with 1.5 L hydration fluid during one hour (2.5% glucose/0.45% saline at a continuous rate). Subsequently, the subjects received a single dose of 2 ng/kg bodyweight LPS (US standard reference Escherichia coli O:113, NIH Pharmaceutical Development Section, Bethesda, MD, USA) and were then infused with hydration fluid at a constant rate of 150 mL/h. During the endotoxemia experiments, heart rate, blood pressure and the course of LPS-induced symptoms such as fever, muscle aches and nausea were constantly monitored.

Blood samples were obtained from 15 multitrauma patients. Trauma patients enrolled in this study, were part of a clinical trial performed at the UMCU. The study was approved by the local ethics committee (ClinicalTrials.gov number NCT03489577, ABR 43279). Written informed consent was obtained from all patients or their legal representatives in accordance with the Helsinki Declaration. Patients suffering from multitrauma who were admitted to the intensive care unit (ICU) of the University Medical Center Utrecht with an expected ICU stay of at least 48 hours were included in this study. Exclusion criteria were: <18 or >80 years old, an altered immunological status and pregnancy. In Table 1 the baseline characteristics of included patients are displayed. Besides the 15 included patients, two additional patients were eligible for inclusion but no informed consent could be obtained and therefore had to be excluded. When patients met the inclusion criteria, the first blood sample was obtained as soon as possible and at least within 12 hours after hospital admission. Subsequent blood samples were obtained at day 3, 6, 10 and 14 or 15 after trauma. Blood samples were drawn in 4 mL sodium heparin tubes. Some patients have missing data points. Missing data is caused by patients either leaving the hospital within the study period or because of in hospital death within the 15 following days after trauma. Relevant clinical data was extracted from the patient files.

The 41 included COVID-19 patients and the bacterial infection patient were part of a prospective cohort study conducted in the University Medical Center Utrecht (UMCU, Utrecht, the Netherlands) during the first wave of the COVID-19 pandemic as described in detail elsewhere by Spijkerman et al. (17). For this study, a waiver for formal ethical approval was provided by the institutional medical ethics committee under protocol number 20-284/C. In short, COVID-19 suspected patients were included upon presentation at the emergency department, where also the first 4 mL blood sample was taken in a sodium heparin tube. All patients were tested for the presence of virus by SARS-CoV-2-specific PCR. Patients were categorized as “COVID-19 positive” if PCR results came back positive. For the specific sub-analysis of pre- and post ICU patients in this study, only male and (non-pregnant) female patients were selected from the database, who had at least one sample taken on the COVID-19 ward <7 days after ICU discharge. Nine of these patients also had data from pre-ICU measurements that were analyzed. Unfortunately, no samples could be obtained during ICU stay. Five patients were excluded from the analysis because they did not have a post-ICU measurement within 7 days after ICU discharge or due to in-hospital death. Relevant clinical data was extracted from the patient files. For baseline characteristics, see Table 1 .

Among patients included in this cohort who eventually tested negative for COVID-19, some patients tested positive for a bacterial infection based on microbiological cultures. One of these patients was selected as example for acute bacterial infection. This specific patient had a clear onset of symptoms of urinary tract infection, a positive microbial culture and a blood sample taken on the same day. This was an exceptional example of the first phase of onset of a bacterial infection.

For the experimental endotoxemia volunteers’ samples, the erythrocytes in blood- and bone marrow samples were lysed using isotonic ice-cold lysis buffer (150 mM NH₄Cl, 10 mM KHCO₃ and 0.1 mM Na₂EDTA dissolved in H₂O; pH of 7.4). Next, leukocytes were washed and resuspended in FACS staining buffer (4 mg/ml human albumin [Sanquin, Amsterdam, The Netherlands] and 0.32% (w/v) sodium citrate in PBS). Blood- and bone marrow samples were stained with a combination of 10 monoclonal antibodies, fixed in 1% PFA and measured on a BD-LSR Fortessa flow cytometer (Becton Dickenson, Mountain View, CA). The following antibody/fluorochrome conjugations were used: anti-CD35-FITC (clone E11), anti-CD64-APC (clone 10,1), anti-CBRM1/5-Alexa Fluor 700 (clone CBRM1/5), anti-CD11b-APC-Alexa Fluor 750 (clone Bear1), anti-CD305 (LAIR-1)-PE (clone DX26), anti-CD14-eF450 (clone 61D3), anti-CD16-Krome Orange (clone 3G8), anti-CD62L-BV650 (clone DREG 56), anti-CD49d-PECy7 (clone G9F10) and anti-CD66b-PerCPCy5.5 (clone G10F5).

Blood samples of trauma patients, (suspected) COVID-19 patients and healthy controls were measured on an AQUIOS CL^® “Load & Go” flow cytometer (Beckman Coulter, Miami, FL, USA). The detailed methods are described elsewhere by Spijkerman et al. (17, 19). For the trauma patients, antibody mixes were used containing anti-CD16-Krome Orange (clone 3G8), anti-CD62L-ECD (clone DREG56), CD10-PC7 (clone ALB1), CD35-FITC (clone J3.D3), anti-CD11c-PeCy5.5 (clone BU15), anti-CD66b-PerCPCy5.5 (clone G10F5) and anti-active CD11b-Alexa700 (clone CBRM1/5). In the COVID-19 cohort and for the healthy controls the following antibody panel was used: CD16-FITC (clone 3G8), CD11b-PE (clone Bear1), CD62L-ECD (clone DREG56), CD10-PC5 (clone ALB1) and CD64-PC7 (clone 22).

Some of the trauma- and COVID-19 samples were also sorted on a FACSAria III flow cytometer (Becton Dickenson, Mountain View, CA, USA) for subsets based on CD16 and CD62L expression, to check for morphological characteristics of the nuclei. Antibody conjugates against CD16-PE-Cy7 (clone 3G8) and CD62L-PErCP-Cy5.5 (clone DREC56) were used.

Cytospin slides of the sorted neutrophil populations were prepared and stained with May-Grünwald-Giemsa staining to determine the nuclear lobularity. The nuclear lobe count was determined by manual counting after visualization on an Axioskop 40 microscope (Zeiss, Jena, Germany) with a 40 objective or a 100x oil immersion objective. Nuclear lobes were considered to be separated if the connection between lobes was less than one third of the width of the adjacent lobes. Progenitors could be identified by nuclear morphology as well as a blue cytoplasm and were counted as having one nuclear lobe (20).

Graphpad Prism version 8.3.0 (GraphPad Software LLC, San Diego, CA, USA) was used to analyze data. Data is presented in graphs as individual data points with mean +/- SD and for the longitudinal trauma data as individual data points with medians. For the experimental endotoxemia data, a one-way ANOVA analysis (paired) with a post-hoc Dunnett’s multiple comparisons test was used to compare blood counts of subsets at different time points (baseline, 1^st and 2^nd challenge). For the bone marrow samples, a two-way ANOVA analysis (paired) with a post-hoc Sidak’s multiple comparisons test was used to compare the presence of multiple subsets in bone marrow at baseline and after the second LPS challenge. For trauma (HC vs day 0, HC vs day 10) and COVID-19 (HC vs pre-ICU, HC vs post-ICU) the same analysis was done as described for the endotoxemia blood samples, although unpaired. Statistical significance was accepted at P^* ≤ 0.05, P^** ≤ 0.01, P^***≤ 0.001 or P**** <0.0001.

Materials and methods concerning the data on CD11b in neutrophil subsets in post-ICU COVID-19 patients and trauma patients and the data of COPD and HIV patients, serving as reference for chronic inflammatory conditions, are discussed in the Supplemental Material .

Circulatory neutrophils from healthy controls (HC) mainly displayed uniformly high expression of CD16 and CD62L ( Figures 1A “Baseline” and 1B ). CD16^low neutrophils were virtually absent (mean counts: 0.09*10⁶ cells/mL, SD: 0.06*10⁶ cells/mL), whereas CD62L^low counts were low but variably present between individuals (mean counts: 1.00 *10⁶ cells/mL, SD: 0.77*10⁶ cells/mL). After LPS administration, CD16^low counts (mean: 2.98*10⁶ cells/mL, SD: 0.99*10⁶ cells/mL, P= 0.0007) and CD62L^low counts (mean: 2.94*10⁶ cells/mL, SD: 1.28*10⁶ cells/mL, P=0.045) were significantly higher in peripheral blood ( Figures 1A “First challenge” and 1B ) (21). CD16^low neutrophils displayed characteristic immature banded nuclear morphology, whereas nuclei of CD62L^low neutrophils were hypersegemented ( Figure 1A “First challenge”) (21).

Trauma patients at the day of hospital admission (day 0) were characterized by the presence of CD16^low neutrophils in the bloodstream within hours after injury (mean: 5.18*10⁶ cells/mL, SD: 2.90*10⁶ cells/mL, P<0.0001) when compared to HC (mean: 0.09*10⁶ cells/mL, SD: 0.08*10⁶ cells/mL). CD62L^low neutrophil counts were somewhat higher early after injury compared to HC, but this did not reach statistical significance (mean 0,80*10⁶ cells/mL, SD: 0.52*10⁶ cells/mL, P=0.67) ( Figures 2A, C ). Furthermore, in the blood of a patient with a diagnosed bacterial infection early after hospitalization, a similar composition of CD16^low and CD62L^low neutrophils was recruited to the circulation in this acute phase after onset of infection ( Figure 2B ).

The high numbers of CD16^low neutrophils found in the circulation early after trauma were not found during the following days ( Figures 2A and 3A, B upper right panel). The CD16^low counts 10 days after trauma were not significantly increased in the circulation compared to HC (mean: 1.23*10⁶ cells/mL versus 0.09*10⁶ cells/mL for HC, P=0.13). CD62L^low neutrophil counts, on the other hand, were significantly elevated after 10 days (mean: 2.03*10⁶ cells/mL versus 0.59*10⁶ cells/mL for HC, P<0.0001) ( Figure 2B ).

Next, we followed the kinetics of the different neutrophil subsets in developing inflammation. The same multitrauma patients were followed in time and their neutrophil subsets were determined at day 0 (right after admission) or day 0,5 (several hours after admission), 3, 6, 10 and day 14 or 15 after admission. As can be seen in Figure 3B (upper left panel), total neutrophil counts were above the normal range early after trauma (day 0) and again late after trauma (at days 10–15) (22). Neutrophilia at day 0 is mainly caused by high counts of CD16^low neutrophils and decreased quickly within hours, as can be seen in the same graph (day 0 vs day 0,5 and day 3). Interestingly, within 3 days after multitrauma, the CD16^low population virtually disappeared from the circulation. The CD62L^low population on the other hand, started increasing from day 3 onward and reached its peak at day 10 ( Figures 3A, B , upper- and lower right panels). This increase in CD62L^low neutrophils in combination with a surge in CD16^highCD62L^high neutrophils were responsible for the second phase of neutrophilia.

CD16^low neutrophils showed a clear band-shaped nucleus. Neutrophils found at later time points (i.e. day 10) were harder to examine on cytospin slides, because they were more fragile ( Figure 3A ). Nevertheless, CD62L^low cells that could be observed indeed expressed more lobes than mature CD16^highCD62L^high cells. In addition, the presence of vacuoles in the nucleus and cytoplasm was noteworthy. This was also seen in CD62L^low neutrophils at other time points after trauma (data not shown).

A similar situation to trauma at day 10 was found in post-ICU patients suffering from COVID-19 ( Figure 4 ). These patients who returned from the ICU also exhibited a pronounced increase in circulating CD62L^low neutrophils (mean: 2.22*10⁶ cells/mL, SD: 1.24*10⁶ cells/mL, P<0.0001). The CD62L^low cells found in these post-ICU patients showed hypersegmentation and fragility ( Figure 4A , second panel). CD62L^low cells were not significantly present before ICU admission (mean: 0.99*10⁶ cells/mL, SD: 1,04*10⁶ cells/mL, P=0.54). In COVID-19 patients CD16^low neutrophils were not increased pre-ICU (mean: 0.07*10⁶ cells/mL, SD: 0.05*10⁶ cells/mL, P=0.85) nor were they post-ICU (mean: 0.12*10⁶ cells/mL, SD: 0.15*10⁶ cells/mL, P=0.71) compared to healthy controls.

We also checked for the activation status of CD62L^low neutrophils by measuring de median fluorescent intensity (MFI) of CD11b in the different subsets of trauma patients at day 0 and 10 and post-ICU COVID-19 patients. Compared to the mature CD16^highCD62L^high subset (mean MFI: 315444, SD: 140583), CD62L^low neutrophils were higher in CD11b in trauma patients at day 0 (mean MFI: 611250, SD: 219799, P= 0,027). This was also the case in trauma patients at day 10 when comparing mature CD16^highCD62L^high neutrophils (mean MFI: 254296, SD: 161328) to CD62L^low cells (mean MFI: 286422, SD: 147249, P=0.037). Results are shown in Supplemental Figure 2A . In post-ICU COVID-19 patients, the increase in CD11b expression of CD62L^low neutrophils (mean MFI: 2262721, SD: 699731) was also significant compared to CD16^highCD62L^high neutrophils (mean MFI: 1720767, SD: 681105, P <0,0001, Supplemental Figure 2B ).

In healthy bone marrow, neutrophil progenitors were present, consisting of promyelocytes, myelocytes and metamyelocytes (data not shown) (23) and also mature CD16^highCD62L^high neutrophils were present. In addition, both CD16^low neutrophils and CD62L^low neutrophils could be found in bone marrow during homeostasis ( Figure 1C “Baseline”). We scored the lobularity of CD62L^low neutrophils in the bone marrow and compared this to CD62L^low neutrophils from the circulation after LPS administration. In contrast to the increased lobularity found in CD62L^low cells from the circulation after LPS administration, trauma and COVID-19 infection, the lobularity of CD62L^low neutrophils in the bone marrow overlapped with those of normal mature CD16^highCD62^high neutrophils in bone marrow ( Figures 1D, E ).

As a model system for repeated inflammatory insults, we monitored neutrophil subsets after a second LPS challenge seven days after the first. Four hours after the second LPS challenge the neutrophil compartment was remarkably similar to baseline samples. Neither CD16^low (mean: 0.27*10⁶ cells/mL, SD: 0.29*10⁶ cells/mL, P=0.208) nor CD62L^low neutrophil subsets (mean: 1.50*10⁶ cells/mL, SD: 0.69*10⁶ cells/mL, P=0.308) were significantly recruited to the circulation when compared to baseline ( Figure 1A “Second challenge” and 1B ) (24). We also monitored the bone marrow four hours after the second LPS challenge and found CD16^low neutrophils were present in the bone marrow in the same proportions as during homeostasis ( Figure 1C ). These data suggest CD16^low neutrophils fail to be recruited from the bone marrow to the circulation during this second inflammatory stimulus.

In order to study the presence of neutrophil subsets in a low-grade chronic inflammatory condition, we chose to characterize COPD patients suffering from neutrophilic inflammation (25, 26) and therapy naïve HIV infected patients. Similarly as found during the later time-points after multitrauma, hardly any immature CD16^low neutrophils were found in the circulation of COPD ( Supplementary Figure 1A ) and HIV patients ( Supplementary Figure 1B ). CD62L^low neutrophils were similarly low in the circulation of these patients. Overall, the subset profiles of these COPD and HIV patients seemed comparable to those of healthy subjects. The severity of the COPD patients ranged between GOLD I-IV, but no significant differences were found between patients in different GOLD stages (data not shown).