Cytokine signatures of end organ injury in COVID-19

Increasing evidence has shown that Coronavirus disease 19 (COVID-19) severity is driven by a dysregulated immunologic response. We aimed to assess the differences in inflammatory cytokines in COVID-19 patients compared to contemporaneously hospitalized controls and then analyze the relationship between these cytokines and the development of Acute Respiratory Distress Syndrome (ARDS), Acute Kidney Injury (AKI) and mortality. In this cohort study of hospitalized patients, done between March third, 2020 and April first, 2020 at a quaternary referral center in New York City we included adult hospitalized patients with COVID-19 and negative controls. Serum specimens were obtained on the first, second, and third hospital day and cytokines were measured by Luminex. Autopsies of nine cohort patients were examined. We identified 90 COVID-19 patients and 51 controls. Analysis of 48 inflammatory cytokines revealed upregulation of macrophage induced chemokines, T-cell related interleukines and stromal cell producing cytokines in COVID-19 patients compared to the controls. Moreover, distinctive cytokine signatures predicted the development of ARDS, AKI and mortality in COVID-19 patients. Specifically, macrophage-associated cytokines predicted ARDS, T cell immunity related cytokines predicted AKI and mortality was associated with cytokines of activated immune pathways, of which IL-13 was universally correlated with ARDS, AKI and mortality. Histopathological examination of the autopsies showed diffuse alveolar damage with significant mononuclear inflammatory cell infiltration. Additionally, the kidneys demonstrated glomerular sclerosis, tubulointerstitial lymphocyte infiltration and cortical and medullary atrophy. These patterns of cytokine expression offer insight into the pathogenesis of COVID-19 disease, its severity, and subsequent lung and kidney injury suggesting more targeted treatment strategies.

Since the outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in December 2019, more than 25 million have developed Coronavirus disease ,with greater than 840,000 deaths 1 . Although patient characteristics vary by geographic location and pandemic stage, underlying conditions such as obesity, hypertension, chronic obstructive pulmonary disease, and diabetes mellitus are consistent risk factors for severe pneumonia [2][3][4][5] .
In addition to pneumonia, COVID-19 patients are at high risk of developing multiorgan systemic complications, including acute respiratory distress syndrome (ARDS), myocardial dysfunction, thrombosis, and acute kidney injury (AKI) 6,7 . In COVID-19 patients who require hospitalization, ARDS occurs in 14% of patients and AKI occurs in 6-9% 8,9 . In intensive care unit (ICU) cohorts, ARDS and AKI are even more common, affecting 73% and 43%, respectively 7 . These complications contribute to the high in-hospital mortality of COVID-19 patients. Although mortality rates vary by location, the latest data shows an overall in-hospital mortality rate of 10% 3,7 .
Increasing evidence shows COVID-19 disease progression and severity may be driven by a dysregulated immunologic response due to over-activation of innate immune pathways, which results in the release of inflammatory cytokines and chemokines, and a corresponding depletion of several lymphocyte populations [10][11][12][13][14] .
Scientific Reports | (2021) 11:12606 | https://doi.org/10.1038/s41598-021-91859-z www.nature.com/scientificreports/ Overproduction of proinflammatory cytokines such as interleukin (IL)-1α, IL-1β, IL-6, IL-10, and tumor necrosis factor-α (TNF-α) have been described in multiple studies compared to healthy controls 15,16 . Despite these reports, there is little data comparing the cytokine profiles of confirmed COVID-19 patients to control patients who present to a hospital in the same time period with symptoms closely resembling COVID-19 but a negative PCRtest. It is also unclear how cytokine expression correlates with clinical parameters and evolves early in the course of an admission to the hospital. In addition, it remains unknown whether specific patterns of dysregulated cytokines are associated with the development of distinct organ dysfunction such as ARDS and AKI in COVID- 19. Identifying potentially diverging inflammatory pathobiology in specific organ dysfunction could suggest differential avenues of treatment. Therefore, the main objective of our study was to assess differences in inflammatory cytokines in COVID-19 patients compared to contemporaneously hospitalized controls, and then to analyze the relationship between these cytokines and the development of mortality, ARDS and AKI.

Results
Demographic and baseline characteristics. A  Initial organ failure, respiratory support and clinical outcomes. Differences in baseline severity of illness were evaluated using admission burden of organ failure, patterns of chest imaging and initial level of oxygen. Despite requiring similar overall levels of supplemental oxygen at admission (p = 0.2, Table 3), 47% of the COVID-19 group were treated with any oxygen compared to 29% of the control population. Day 1 of hospital admission SOFA scores were higher in COVID-19 patients when compared to controls (2.0 [1.0, 5.0] vs 1.0 [0.0-3.0]; p = 0.001). Additionally, chest X-ray findings upon arrival to the emergency room were different between groups. The majority of COVID-19 patients had bilateral infiltrates at admission compared to controls (64% vs 13%; p < 0.001).
To evaluate the relative in-patient morbidity and mortality, we followed the COVID-19 and control patients through their index hospitalization and documented incident complications to compare differences between the groups. COVID-19 patients more commonly developed ARDS (40% vs 0%; p < 0.001) as well as any kidney injury as shown in Table 3, including treatment with kidney replacement therapy (KRT) (14% vs 2%, p = 0.037) compared to controls. The 28-day and in-hospital mortality in the COVID-19 compared to the control groups were 19% vs 8% p = 0.13 and 21% vs 9.8%; p = 0.14 respectively. Inflammatory cytokine expression in COVID-19 compared to controls. Our data thus far demonstrated that compared to controls, the baseline severity of illness was higher in the COVID-19 group and that they frequently developed in-patient complications. To explore whether these findings were related to differences in inflammatory cytokine expression, we analyzed the day 1 serum cytokine profile by 48plex. Several differences between the COVID-19 and control day 1 of hospital admission cytokine expression levels were identified. Specifically, there was a significant overexpression of IP-10, TNF-α, IFN-α2, IFN-γ, IL-1RA, MCP-3, M-CSF, IL-7, MCP-1, MIP-1β, IL-15, IL-12 (p40), PDGF AA, IL-6, FLT 3L, and IL-10 in COVID-19 patients, as shown in Fig. 1. The log 2 fold-change differences between groups are shown in Supplementary Table 5. To expand these findings and identify whether there was dose-response relationship between overexpressed cytokines and the severity of COVID-19 pneumonia by the WHO classification, we examined whether cytokine expression levels associated with disease severity. When compared to mild disease, patients that developed severe pneumonia had a significant increase in IL1-RA, IL-6, IP-10, MCP-1, MCP-3, M-CSF, and TNF-α (Fig. 2).
Cytokine correlation with clinical laboratory findings. We next explored the relationship between the 48plex cytokines and routinely measured clinical laboratory values to understand how closely the baseline clinical and inflammatory phenotypes were correlated within COVID-19 and control patients. As shown in Cytokine correlation with baseline organ dysfunction and other cytokines. We next examined the correlation between cytokine levels and admission SOFA score among COVID-19 and control patients to explore relationships with common clinical phenotypes of acute organ dysfunction. As illustrated in Fig. 4 Supplementary Fig. 3, there were differential correlations between cytokine levels and SOFA scores in the two   Supplementary Fig. 3). Given the differential relationship between cytokines and clinical parameters in COVID-19 patients and controls we further analyzed the relationship between the cytokines in each group (Fig. 4). There were more cytokine to cytokine correlations in the COVID-19 group compared to controls. Additionally, distinguishable inflammatory cytokines and chemokines were positively correlated within COVID-19 patients, such as: macrophage induced chemokines (IL-1β, IL-1RA, IL-6, IL-12, IL-18, CXCL1/GROα, CCL7/MCP3, CCL2/MDC, CCL3/MIP-1α, TGF-α, TNF-β and IFN-α2), T-cell related interleukines (IL-4, IL-5, IL-13, IL-15, IL-17A, IL-17E/ IL-25 and sCD40L) and stromal cell producing cytokine (IL-7).

Cytokine correlation with clinical outcomes.
Our data demonstrated a differential inflammatory phenotype in COVID-19 compared to controls with unique relationships between baseline cytokines, clinical labs, and organ dysfunction. We next evaluated whether differential baseline cytokine levels in the COVID-19 group associated with the development of subsequent clinical outcomes of ARDS, AKI and mortality. In the COVID-19 cohort, thirty-six patients developed ARDS (40%) ( Table 3). The Cox Proportional Hazard (PH) regression model results for cytokines associated with ARDS within COVID-19 patients are shown in Fig. 5 Table 3). The maximum AKI staging developed during hospitalization for COVID-19 patients was 13% for stage 1 AKI, 5.6% for stage 2 AKI, and 25% for Stage 3 AKI (Table 3). As shown in Supplementary Fig. 4, there was a significant association with IL-6 levels and AKI staging (p = 0.009) and relevant trends with TNF-α, IL-1RA, and FGF-2. The Cox PH regression model results for development of AKI within COVID-19 patients are shown in Fig. 5 Table 8b). Of note, IL-13, secreted by activated Th2 cells, constituting a counter-regulatory system for the inflammatory response was not only correlated mortality but also predictive of ARDS and AKI.
Longitudinal changes in cytokine profile in the COVID-19 group. Potential longitudinal changes of cytokine levels throughout early days of the hospital were explored in 15 COVID-19 patients with cytokine measures for each of the first three days following admission to the general ward floor. There were no detectable trends for any of the cytokines over this time period (Supplementary Fig. 2, and Supplementary Tables 1 and 2). We validated this result using an 8-plex assay in a subset of 54 COVID-19 patients (Supplementary Table 4). Consistent with the findings in 48-plex, there were no detectable differences in cytokine expression within 3 days post admission.

Histopathology findings of COVID-19 lung and kidney.
We have demonstrated that higher levels of inflammatory cytokines and chemokines during COVID-19 infection are associated with disease severity and death. Since organ-specific injury such as lung and kidney 17 may be a contributing factor, we examined histopathological features of lung and kidney specimen from COVID-19 patients in the cohort who died dur- Table 3. Severity of Illness and outcomes of COVID-19 patients and controls. AKI was defined according to KDIGO guidelines 21 . ARDS was defined according to the Berlin Definition 22 . HFNC High-flow nasal cannula, NRB non-rebreather mask, NIV non-invasive, CXR chest X-ray, SOFA sequential organ failure assessment. a Patients with end stage renal disease who were on dialysis prior to admission were excluded.  (Fig. 6a). Additionally, the lungs of COVID-19 patients demonstrate greater epithelial cell injury in comparison to controls, indicated by an increase in the number of terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) positive cells. In COVID-19 kidneys, progressive glomerular sclerosis, tubulointerstitial lymphocyte infiltration and moderate to severe cortical and medullary atrophy were observed (Fig. 6b). As with lung, TUNEL positive cells were increased in periglomerular and tubular epithelial cells of COVID-19 patients, suggesting that COVID-19 infection may cause glomerular and renal tubular injury (Fig. 6b). The pathologic features of cortex and tubulointerstitium in COVID-19 kidney and controls in each Banff scoring is shown in Fig. 6b. Total inflammation (ti) was elevated in COVID-19 www.nature.com/scientificreports/ patients compared to controls. Chronic changes such as tubular atrophy (ct) and interstitial fibrosis (ci) were also significantly increased in COVID-19 cases (Fig. 6b).

Discussion
This study highlights important cytokine expression differences between COVID-19 patients and similarly presenting, contemporaneously admitted control patients. Within the COVID-19 cohort, we found numerous correlations with the baseline burden of organ dysfunction as well as with the subsequent development of critical clinical outcomes of ARDS, AKI, and death. We revealed patterns of dysregulated cytokines associated with the presence of COVID-19 disease, its severity, and subsequent lung and kidney injury 6,18 . However, we did not find substantial variation in our cytokine panel across the first three days in our COVID-19 group. www.nature.com/scientificreports/ In our analysis of admission cytokines and baseline organ injury, several Th1 and Th2 associated cytokines were correlated to SOFA score in COVID-19 patients as shown in Fig. 4. This suggests potential imbalances in Th1 and Th2 driven immunity, which could lead to altered neutrophil recruitment, monocyte and epithelial activation. This in turn supports the increased mononuclear inflammatory infiltrates in the histopathology findings present in fatal COVID-19 cases 19 . These findings, in combination with the apparent stability in the cytokine panel over time, suggest that the proposed pathophysiological response is persistent. We associated cytokine signatures at admission with histopathological findings from the same patients that ultimately succumbed to COVID-19 during their index admission. Our autopsy findings of COVID-19 patients demonstrate mononuclear inflammatory cell infiltration such as macrophages and lymphocytes in both lung and kidney tissues. This is consistent with the findings of increased macrophage and lymphocyte derived cytokines in plasma samples of COVID-19 patients compared to controls.
Given the accumulation of evidence of the systemic effects of COVID-19, we analyzed a broad cytokine panel's relationship with the subsequent development of ARDS, AKI, and death. ARDS and AKI in COVID-19 are associated with a high morbidity even without mortality 6,7,9,20 and the role of cytokines in the development of these complications remains poorly understood. Our data shows that inflammatory cytokine signatures are associated with systemic disease severity beyond pneumonia. Specifically, the development of both ARDS and AKI in COVID-19 patients is associated with IL-1RA, fractalkine, M-CSF, G-CSF, IL-6, and TNF-α, and supports the potential deleterious effect of the "cytokine storm" on disease progression 15,21 . IL-1RA, which is mostly produced by epithelial cells, can also be made by multiple types of immune cells and by binding to the IL-1R can act as a natural inhibitor of IL-1ß. Given the association with both ARDS and AKI, it is plausible that IL-1RA is potentially associated with changes in pulmonary function, lung damage, and increased kidney injury. Interestingly, IL-1RA plays an important role in lipid metabolism, fever generation, neutrophil chemotaxis, positive www.nature.com/scientificreports/ regulation of IL-6 production, and the acute-phase response of infection. As epithelial cells and macrophages are the main producers of IL-6 in the lung, elevated IL-1RA expression may further augment IL-6 production by these cells; thereby, contributing to a deleterious positive feedback loop within COVID-19 patients that can directly impact both the lung and kidney. Further, as demonstrated by our results, heightened TNF-α production was indicative of severe disease progression in COVID-19 patients. As there is an inverse relationship between TNF-α expression and T cell recruitment, similar to IL-1RA and IL-6, elevated TNF-α expression may further contribute to inflammatory progression of COVID-19 patients to develop ARDS or AKI. However, increased IL-18 on day 1, a marker for inflammasome activation, was inversely associated with mortality. Appropriate inflammasome activation may play a critical role in the host defense during SARS-CoV-2 infection 22 . Pyropstosis, an inflammasome and casapase-1 mediated programmed cell death, has been shown to play an important role in viral diseases 23 . Inflammasome recognition of viral molecules can lead to the activation of pyroptosis by promoting caspase-1 activation and IL-1β and IL-18 23 . Additionally, Inflammasome impairment have been shown to decrease survival in elderly mice during Influenza infection 24 .
Our study also demonstrated levels of chemokines, such as G-CSF and M-CSF, correlated with the development of ARDS and AKI in COVID-19 patients, suggesting a role of leukocyte maturation and activation in disease progression. M-CSF is a primary chemokine associated with the growth, proliferation, and differentiation of hematopoetic cells, including monoblasts, pro-monocytes, monocytes, macrophages, and osteoclasts. M-CSF is secreted by monocytes, fibroblasts, stromal cells, and endothelial cells 25 . As demonstrated in our current findings, higher levels of M-CSF were not only associated with pneumonia severity, but also highlighted the potential for monocyte/macrophage driven development of either ARDS or AKI. GCSF is produced by macrophages and the endothelium and is essential for the proliferation and maturation of neutrophils, eosinophils, and basophils. In response to elevated G-CSF, proliferation and differentiation of precursor cells into mature granulocytes occurs 26 . Mature granulocytes, play an important role in chemotaxis, phagocytosis, as well as the release of lysosomal enzymes at sites of infection. Increased G-CSF can occur in patients with neutropenia as a feedback mechanism to increase neutrophil migration to the site of infection. While these cells are essential for the host antiviral response, an overzealous or heightened response can lead to increased cellular death and loss of homeostasis. Based on our current findings, we hypothesize that the cellular dysfunction and death in the lung in severe COVID-19 modifies the initiation and regulation of an 'effective' innate immune response. Importantly, as these cytokine signatures are present in plasma, it is plausible that this dysregulated, overly zealous immune response can result in systemic organ dysfunction.
We found that MCP-1, MCP-3, IP-10, and IL-8, which associated with ARDS, were not associated with AKI and mortality, suggesting a differential role in monocyte migration and macrophage activation in disease development. MCP-1 is a powerful monocyte chemotactic factor that is constitutively produced by oxidative stress, cytokines, or growth factors and can be expressed by endothelial cells, fibroblasts, epithelial cells, monocytes, and macrophages. Similar to M-CSF, MCP-1 plays an important role in the antiviral response and regulates the migration and infiltration of monocytes and NK cells 27 . MCP-3 is produced by macrophages and attracts monocytes to The final injury score was derived from the following calculation: Score = I + ii + iii + iv + v + vi + vii + viii. G indicates glomerulus. Banff Score: g, glomerulitis; i, interstitial inflammation; ptc, peritubular capillaritis; ct, tubular atrophy; ci, interstitial fibrosis; cv, vascular fibrous intimal thickening; ti, total inflammation. *p < 0.05, **p < 0.01. www.nature.com/scientificreports/ the site of infection 28 . MCP-3 regulates macrophage function through its binding to chemokine receptors CCR1, CCR2, and CCR3. IL-8 can act as a chemoattractant to recruit neutrophils and other immune cells to the site of infection. IL-8 is secreted by macrophages, but can also be released by epithelial cells, airway smooth muscle cells, and endothelial cells. Of note, IL-8 is involved in multiple cellular processes, such as tissue proliferation, tissue remodeling, and angiogenesis 29 . IP-10 is secreted by neutrophils, endothelial cells, fibroblasts, dendritic cells, and hepatocytes. IP-10 binds to CXCR3 to regulate immune system responses by activating and recruiting leukocytes, including T cells, monocytes, and NK cells. Recruitment of leukocytes to inflamed tissues can perpetuate inflammation, and thereby, increased IP-10 can contribute to extensive tissue damage 30 . In summary, as detailed by our data, heightened expression of MCP-1, MCP-3, IL-8, and/or IP-10 demonstrates how an overzealous monocyte/macrophage driven immune response can contribute to ARDS development in COVID-19.
Although prior studies have noted the association of increased pro-inflammatory cytokines in COVID-19 pneumonia severity [10][11][12][13]31 , we found distinct cytokine signatures for eventual ARDS, AKI and mortality. In contrast to mortality, we observed that ARDS and AKI development were associated with macrophage migration, and immune cell and epithelial activation, which suggests an exclusive Th1 driven immunity by TNF-β and IFN-α2.
We saw less correlations with between novel cytokines, clinical labs, and severity of illness in our control patients. This result supports a more homogeneous but graded dysregulated immune response in COVID-19. The increased proportion of patients with cancer and immunosuppression in the control population does yield potential to bias our results, however, cancer typically elicits an inflammatory response. Moreover, We did not explore cytokines within urine samples in this study. Additionally, it is important to mention that standard treatment for COVID-19 at the time of our sample collection did not include routine use of dexamethasone or remdesivir which may alter the relationship between cytokines and eventual outcomes. Moreover, the surge conditions in New York City may have affected the clinical practice pattern during the study period. However, our in-hospital mortality was similar to other reports from around the United States and the world 3,7 .
Overall, our findings support the role of dysregulate broad cytokine response in the pathogenesis of severe COVID-19. Specifically, our study provides an extensive analysis on how cytokine signatures differentially associate with baseline organ failure, clinical labs, eventual disease severity, and organ specific complications of ARDS and AKI. These findings highlight potential leukocyte and monocyte specific immunologic signatures that suggest differential pathophysiology leading to AKI, ARDS, and mortality. Cytokine signatures could be used for the prognostication of specific types of organ dysfunction. They also suggest alternative therapies for patients with COVID-19.

Methods
Human subjects and design. Our cohort study included adults 18 years of age or older with confirmed COVID-19 and a population of SARS-CoV-2 negative controls who were admitted to the general wards between March 3, 2020 (date of the first case) and April 1, 2020 at an 862-bed quaternary referral center in New York City. All patients presented to the emergency department for an acute complaint and were subsequently admitted for inpatient care. COVID-19 cases had presenting symptoms, fever, cough, and dyspnea consistent with COVID-19 and were confirmed through reverse transcriptase polymerase chain reaction (rt-PCR) assay for SARS-CoV-2, performed on nasopharyngeal swab specimens. Control patients were made up of contemporaneous admissions to the hospital who had a negative rt-PCR for SARS-CoV-2 and were not considered to have an illness consistent with COVID-19. Pediatric and pregnant patients were excluded from the study. Additionally, autopsy specimen, if available, from patients in the cohort were included. The study was approved by the institutional review board of Weill Cornell Medicine (20-05022072, 19-10020914, 20-04021880, 20-04021796, and 20-03021681) with a waiver of informed consent for serum specimens and autopsy consent was obtained by next of kin which includes informed consent for study participation. All methods were performed in accordance with the relevant guidelines and regulations.
Clinical evaluation. Baseline demographics, clinical characteristics, comorbid conditions, vital signs, laboratory values, and radiographic findings on presentation were manually abstracted from the electronic health record by trained research personnel with the use of a quality-controlled protocol and structured abstraction tool 4 . Laboratory and radiographic testing were performed according to clinical needs and analyzed/interpreted on site. The Sequential Organ Failure Assessment (SOFA) score, a severity of illness score that sums six separate organ dysfunction subscores, was used to characterize baseline severity of illness. For the central nervous system, kidney, liver, and coagulation organ dysfunction subscores, traditional SOFA methodology 32 was used. When the respiratory SOFA subscore was not available due to a lack of partial pressure of oxygen (PaO 2 ), we used a commonly accepted imputation technique to impute PaO 2 from an oxygen saturation (SpO 2 ) level 33 . The cardiovascular SOFA subscore was updated with additional vasopressors according to a norepinephrine equivalency table 34 . Missing data for each subscore was treated as normal. We additionally classified COVID-19 patients as moderate or severe based on the World Health Organization (WHO) interim guidelines system 35 , by in-hospital maximal oxygen and organ failure support. Acute kidney injury (AKI) was defined by the Kidney Disease Improving Global Outcomes (KDIGO) criteria and staging 36 . ARDS was defined according to the Berlin definition 37 as the need for mechanical ventilation, bilateral infiltrates in the chest x-ray, and clinical diagnosis of ARDS by the treating attending physician. Thromboembolic events included any deep vein thrombosis or pulmonary embolism confirmed radiographic imaging. Respiratory co-infections included any other viral, bacterial, or fungal pathogen isolated on any respiratory sample (e.g., nasopharyngeal swab, sputum sample, bronchoalveolar lavage).