Comparison of host endothelial, epithelial and inflammatory response in ICU patients with and without COVID-19: a prospective observational cohort study

Background Analyses of blood biomarkers involved in the host response to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) viral infection can reveal distinct biological pathways and inform development and testing of therapeutics for COVID-19. Our objective was to evaluate host endothelial, epithelial and inflammatory biomarkers in COVID-19. Methods We prospectively enrolled 171 ICU patients, including 78 (46%) patients positive and 93 (54%) negative for SARS-CoV-2 infection from April to September, 2020. We compared 22 plasma biomarkers in blood collected within 24 h and 3 days after ICU admission. Results In critically ill COVID-19 and non-COVID-19 patients, the most common ICU admission diagnoses were respiratory failure or pneumonia, followed by sepsis and other diagnoses. Similar proportions of patients in both groups received invasive mechanical ventilation at the time of study enrollment. COVID-19 and non-COVID-19 patients had similar rates of acute respiratory distress syndrome, severe acute kidney injury, and in-hospital mortality. While concentrations of interleukin 6 and 8 were not different between groups, markers of epithelial cell injury (soluble receptor for advanced glycation end products, sRAGE) and acute phase proteins (serum amyloid A, SAA) were significantly higher in COVID-19 compared to non-COVID-19, adjusting for demographics and APACHE III scores. In contrast, angiopoietin 2:1 (Ang-2:1 ratio) and soluble tumor necrosis factor receptor 1 (sTNFR-1), markers of endothelial dysfunction and inflammation, were significantly lower in COVID-19 (p < 0.002). Ang-2:1 ratio and SAA were associated with mortality only in non-COVID-19 patients. Conclusions These studies demonstrate that, unlike other well-studied causes of critical illness, endothelial dysfunction may not be characteristic of severe COVID-19 early after ICU admission. Pathways resulting in elaboration of acute phase proteins and inducing epithelial cell injury may be promising targets for therapeutics in COVID-19. Supplementary Information The online version contains supplementary material available at 10.1186/s13054-021-03547-z.


Background
Coronavirus disease 2019  is caused by the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) and can result in organ dysfunction including acute respiratory distress syndrome (ARDS), severe acute kidney injury (AKI), and death [1][2][3]. While only a minority of COVID-19 patients will become critically ill, these patients take up a disproportionate share of intensive care unit (ICU) resources and experience high mortality [1,3]. Our previous work has defined markers of endothelial dysfunction as critical determinants of organ failure and mortality among ICU patients, but it is unclear whether COVID-19 shares this pathophysiology [4][5][6]. A clear understanding of the shared and unique characteristics of COVID-19-related critical illness could help to inform the development of novel therapeutics and the repurposing of existing interventions.
Previous reports have suggested that severe COVID-19 is characterized by derangements in inflammatory, endothelial and epithelial cell injury pathways [7]. However, comparisons to other critically ill populations are under-reported. For example, comparisons of plasma biomarkers of endothelial function have largely focused on COVID-19 alone [8], comparisons between COVID-19 and healthy controls [9], and comparisons between patients with increasing severity of COVID-19 [10]. While these studies have advanced our knowledge, the findings are conflicting and the study designs prevent discriminating differences specific to the host response to severe SARS-CoV-2 infection versus a general signature of critical illness. Thus, it is unknown whether inflammatory, endothelial and epithelial cell injury pathways are an early marker of the host response to SARS-CoV-2, and, in turn, if therapies that target these pathways to modulate the host response should be preferentially tested in COVID- 19. To address this knowledge gap, we developed the COVID-19 Host Response and Clinical Outcomes (CHROME) study enrolling a prospective cohort of patients admitted to an ICU as persons under investigation (PUIs) for COVID-19. Blood was collected within 24 h and again 3 days after ICU admission. We hypothesized that biomarkers of the host response would be different early after ICU admission in COVID-19 compared to a matched cohort of ICU patients without COVID-19.

Clinical data and ascertainment of outcomes
We prospectively enrolled 171 ICU patients admitted to three University of Washington Hospitals campuses (Montlake, Harborview and Northwest hospitals) with suspicion of COVID-19 infection, including 78 (46%) patients with a confirmed SARS-CoV-2 infection and 93 (56%) with a negative SARS-CoV-2 test in the CHROME cohort (UW IRB: 9763 and 6878) (Fig. 1). Study enrollment started on April 2nd, 2020 and patients included in this analysis were enrolled until September 14th, 2020. All patients included in this analysis had complete data ascertainment. All patients had symptoms consistent with COVID-19, were placed under respiratory isolation by the treating physician (PUI) and had a nasopharyngeal swab sent for SARS-CoV-2. A confirmed case of COVID-19 was defined by a positive result on a reverse-transcriptase-polymerase-chain-reaction (RT-PCR) assay. The primary aim of our study was to determine whether plasma biomarkers of key pathways in sepsis were different between COVID-19 and non-COVID-19 critically ill patients. Our secondary aim was to determine whether differentially expressed plasma biomarkers were associated with clinical outcomes, such as mortality, AKI and ARDS, in COVID-19 and non-COVID patients.
Clinical data was abstracted from the electronic medical record into standardized case report forms. Data to calculate the APACHE III score was collected in the first 24 h of ICU admission [11]. We also determined patient severity on study enrollment using the 8-point ordinal scale as previously reported [12]. Trained research coordinators identified the initiation of inpatient dialysis (hemodialysis or continuous renal replacement therapy), thrombosis, shock, ARDS and death by chart review and all outcomes were adjudicated until death or hospital discharge. ARDS was categorized by Berlin Criteria [13], reflecting each individual's worst oxygenation level (based on the ratio of the inspired FIO 2 and the arterial PaO 2 ) in combination with the adjudication of chest radiographs by an Attending Radiologist (SP). AKI was defined as an increase ≥ 0.3 mg/dL in 48 h and/or ≥ 50% in serum creatinine concentrations measured during hospitalization compared to a 'baseline' serum creatinine value measured at CHROME study enrollment or need for new dialysis. Severe AKI was defined as doubling of serum creatinine compared to a 'baseline' serum Conclusions: These studies demonstrate that, unlike other well-studied causes of critical illness, endothelial dysfunction may not be characteristic of severe COVID-19 early after ICU admission. Pathways resulting in elaboration of acute phase proteins and inducing epithelial cell injury may be promising targets for therapeutics in COVID-19. Keywords: COVID-19, Critical illness, Acute respiratory distress syndrome, Endothelial dysfunction creatinine value or need for new dialysis after study enrollment [14].

Statistical analyses
For baseline characteristics, we report continuous variables as mean ± SD or median and interquartile range (IQR) and categorical variables as number and percent. Since biomarker concentrations can be skewed all biomarkers were log 2 transformed. Height data was missing for five participants and we used multiple imputation using chained equations, combining across imputations using Rubin's rules. To determine the risk of clinical outcomes, we performed relative risk (RR) regression using a Poisson model and robust standard error estimates to test for associations between clinical outcomes, such as ARDS, severe AKI (Stage 2 or 3), dialysis and death and COVID-19 status. We selected adjustment variables a priori on the basis of biologic plausibility and prior literature suggesting these variables may confound associations between plasma biomarkers and clinical outcomes [17,18]. All comparisons were adjusted for age, sex, body mass index (BMI), APACHE III scores and Charlson comorbidity index.
To compare the fold changes in biomarker concentrations between COVID-19 and non-COVID-19 patients, we used linear regression with robust Huber-White standard errors of the log-transformed plasma biomarker on COVID-19 status. Analyses were adjusted for age, sex, BMI and APACHE III scores. To control type 1 error from multiple hypothesis testing, we used a Bonferroni corrected p value of 2.2 × 10 -3 (0.05/22 biomarkers). We also performed subgroup analyses to explore whether the magnitude of associations between plasma biomarkers and COVID-19 populations differed by primary ICU admission diagnosis. We only carried forward biomarkers that were significantly different using a Bonferronicorrected threshold between COVID-19 and non-COVID-19. We next used a fold ratio of biomarker concentrations from day 1 to day 3 in the subset of patients who had blood collected at both timepoints. We tested the change in fold ratio over time stratified by COVID-19 status.
In secondary analyses, we tested whether plasma biomarkers were differentially associated with clinical outcomes based on COVID-19 status. We used RR regression to identify associations between biomarker concentrations and clinical outcomes stratified by COVID-19 status. We present univariate and multivariate associations between plasma biomarkers and clinical outcomes as RR per doubling of the plasma biomarker. All analyses were performed in R, version 3.6.2 (R Project for Statistical Computing).
The primary ICU admission diagnosis of respiratory failure or pneumonia was approximately 75% in both groups. Causes of respiratory failure in the non-COVID-19 population included bacterial pneumonia, aspiration pneumonia, chronic obstructive pulmonary disease exacerbation and others (Additional file 1: Table S2). Viral causes of pneumonia were uncommon in the non-COVID-19 group. Scores for the 8-point ordinal scale were similar in patients with and without COVID-19 (Table 1). A similar proportion of patients in both COVID-19 and non-COVID-19 groups were receiving oxygen therapy via non-invasive positive pressure ventilation or high flow nasal cannula (17% and score 6) and 46% were receiving invasive mechanical ventilation or extracorporeal membrane oxygenation (score 7) at study enrollment. Mean APACHE III scores were higher in non-COVID-19 (80.8 ± 29.5) than COVID-19 patients (70.5 ± 28.7).

Correlations in key plasma biomarkers reflecting epithelial cell injury, inflammation and endothelial activation/ dysfunction between ICU patients with and without COVID-19
Among non-COVID-19 patients, clinical biomarkers of coagulation (international normalized ratio (INR)) were moderately correlated with endothelial markers (Ang-2 and Ang-2:1 ratio (range of correlation 0.59-0.64). In contrast, in COVID-19 patients, INR was minimally correlated with endothelial markers (range of correlation 0.07-0.16). In non-COVID-19 patients, platelet counts and Ang-1 concentrations were highly correlated (r = 0.72) while the correlation was much weaker in COVID-19 (r = 0.47). In non-COVID-19 patients, inflammatory biomarkers, such as IL-6 and sTNFR-1, were moderate to highly correlated with Ang-2 and Ang-2:1 ratio (range of correlation 0.53-0.71). In contrast, COVID-19 patients these inflammatory biomarkers were minimally correlated with endothelial biomarkers (range of correlation 0.17-0.42). (Fig. 2a, b). These findings suggest the relationship between coagulation and endothelial markers of injury may be distinct in subjects based on COVID-19 status. We also compared 13 clinical biomarkers (serum creatinine, white blood cell count and others) between COVID-19 and non-COVID-19. Among the clinical biomarkers, we found that white blood cell count was significantly lower in COVID-19, while platelet counts were significantly higher in COVID-19 compared to non-COVID-19 (Additional file 1: Table S3).
Since corticosteroids could modulate the host response to infection, we completed a sensitivity analysis adjusting for receipt of corticosteroids. We found that while the associations were mildly attenuated, the same five plasma biomarkers (Ang-2:1 ratio, Ang-2, sTNFR-1, sRAGE, and SAA) remained significantly different between COVID-19 and non-COVID-19 patients (Additional file 1: Figure S1).

Differences in epithelial cell injury and inflammatory pathways between COVID and non-COVID-19
Biomarkers of epithelial cell injury (sRAGE) and acute inflammation (SAA) were higher in COVID-19. Concentrations of sRAGE were approximately twofold greater in

Distinct molecular pathways are associated with clinical outcomes in COVID-19 and non-COVID-19 patients
We next sought to determine whether five plasma biomarkers (Ang-2, Ang-2:1 ratio, SAA, sTNFR-1 and sRAGE) that were present at different levels between COVID-19 and non-COVID-19 patients were associated with clinical outcomes. We found that none of the five biomarkers were associated with hospital mortality in COVID-19 (Table 3). In contrast, higher concentrations of Ang-2:1 ratio were associated with greater risk of hospital mortality only in non-COVID-19 patients. We also found that higher concentrations of SAA were  We then evaluated whether plasma biomarkers were associated with two common forms of organ dysfunction in COVID-19, ARDS and severe AKI. We found that higher sRAGE concentrations were associated with ARDS only in the non-COVID-19 patients, while sRAGE was not associated with ARDS in COVID-19 (Additional file 1: Table S7). In contrast, higher SAA concentrations were only associated with ARDS in COVID-19 (aRR = 1.27 (95% CI 1.10-1.48) but not non-COVID-19 (aRR = 1.02 (95% CI 0.90-1.15). None of the plasma biomarkers were differentially associated with severe AKI based on COVID-19 status. In unadjusted analyses, we found that higher Ang-2 and sTNFR-1 concentrations were associated with a greater risk of developing severe AKI during hospitalization both in COVID-19 and non-COVID-19 populations (Additional file 1: Table S8). The association of Ang-2 with severe AKI was attenuated in both groups after adjusting for APACHE III and Charlson comorbidity index, while the association of sTNFR-1 with severe AKI remained significant in both populations.

Discussion
The pathophysiology of severe COVID-19 has been reported to be characterized by severe inflammation and vascular injury. However, the extent to which these findings are specific to COVID-19 as opposed to critical illness in general is unclear [19]. To our knowledge, this is the first study of a cohort of ICU patients admitted as PUI in which circulating biomarkers have been compared between patients that ultimately ruled in or out for COVID-19. We found distinct differences in the relationship of biomarkers based on COVID-19 status. For example, in non-COVID-19, inflammatory and endothelial biomarkers were highly correlated and consistent with our prior work [5]. In contrast, in COVID-19 inflammatory and endothelial pathways seemed to be uncoupled. Moreover, in COVID-19, there was a lower correlation between coagulation (INR and platelets) and endothelial markers compared to non-COVID-19.
In adjusted comparisons between COVID-19 and non-COVID-19 patients, we found that circulating markers of endothelial dysfunction and inflammation (Ang-2:1 ratio, Ang-2 and sTNFR-1) were lower in COVID-19 compared to non-COVID-19 patients. In contrast, acute phase proteins (SAA), and markers of epithelial cell injury/innate immune activation (sRAGE) were significantly greater in COVID-19, independent of demographics and ICU severity of illness. While mortality rates were similar between groups, dysregulation of Ang-2:1 ratio were only associated with hospital mortality in the non-COVID-19 population. In contrast, SAA was associated with ARDS only in COVID-19 patients. Despite higher sRAGE concentrations in COVID-19 patients, higher sRAGE concentrations were associated with ARDS only in the non-COVID-19 population. These findings suggest that, unlike previously studied causes of critical illness, endothelial dysfunction may not be a distinct signature of the host response to COVID-19 infection, and, in contrast, epithelial cell injury/innate immune activation and inflammation may be a pathway for future study and targeting of therapeutics in COVID- 19.
We found that rates of hospital mortality and organ dysfunction were similar in critically ill patients with and without COVID-19. These findings differ from previous studies reporting extremely high rates of AKI, thrombosis and death in patients with COVID-19 [20]. However, the study designs used in these prior reports comparing COVID-19 clinical outcomes to historical cohorts rather than within a contemporaneous cohort may result in significant confounding. Comparison of clinical outcomes between COVID-19 and a contemporaneous cohort mitigates, though does not eliminate, confounding due to differences in processes of care, differences in practice patterns due to system strains and allocation of limited resources. An additional factor differentiating our study from multiple prior studies is that while shortages of medical resources and ICU beds occurred in a number of other cities and countries, our center did not experience a limitation of resources, staff, or ICU beds and never approached crisis standards of care [21]. In addition, enrollment in the CHROME study occurred from April to September allowing for a broader representation of demographics and comorbidities among both the COVID-19 and non-COVID-19 admissions [22]. In contrast to similarities in mortality and organ dysfunction, COVID-19 patients experienced increased LOS, a very important factor that have strained ICU resources across the globe as a result of the COVID-19 pandemic.
Studies have suggested that COVID-19 is a hyperinflammatory syndrome or "cytokine storm". The presumed elevations in IL-6 concentrations have been hypothesized as causal in severe COVID-19 and this has led to interventional trials testing therapies that block IL-6 [23,24]. Of note, in our study, plasma concentrations of IL-6 and IL-8 were consistently elevated but similar between ICU patients with and without COVID-19 suggesting that there is heterogeneity in the inflammatory response and sub-groups of patients regardless of COVID-19 status may have elevated IL-6 concentrations. Another key inflammatory biomarker measured in our study was sTNFR-1. sTNFR-1 is the circulating form of the membrane-bound receptor which is essential for TNF signaling that results in both inflammation and endothelial dysfunction [25,26] and we and others have shown that sTNFR-1 is highly associated with organ dysfunction and death in sepsis and other forms of ARDS [16]. Counter to what we had hypothesized, we found that sTNFR-1 was lower in COVID-19 relative to non-COVID-19, but our findings are consistent with results from Sinha et al. showing that sTNFR-1 concentrations were lower in patients with COVID-19-associated ARDS compared to a historical cohort of patients with ARDS [27].
In contrast to these prototypical inflammatory cytokines, chemokines, and cytokine receptors we found that the acute phase protein, SAA, was markedly elevated in COVID-19. Our results extend a prior report from China that demonstrated that SAA concentrations were associated with severity of respiratory failure and mortality in hospitalized patients with COVID-19 [28]. Finally, levels of sRAGE, a marker of alveolar type 1 cell-injury that is associated with risk for ARDS and related mortality, were significantly greater in COVID-19 patients [29,30]. Our findings suggest that SAA and sRAGE may be useful indications of COVID-19 pathophysiology.
Considerable interest has focused on the potential role for vascular pathology in the development of severe COVID-19 [7]. However, in our survey of multiple markers of endothelial function/dysfunction and injury, we only observed significantly lower Ang-2 and Ang-2:1 ratio in COVID-19 relative to non-COVID-19 patients. We also found a trend toward lower sTie-2 and higher concentrations of Ang-1 in COVID-19. Previous studies have demonstrated that Ang-1 and 2 concentrations are associated with mortality in community acquired pneumonia and sepsis [6,31]. However, the majority of studies examining Ang-1 and Ang-2 in critical illness were likely to have had had an underlying bacterial (as opposed to viral) etiology [31,32]. Our results suggest that endothelial dysfunction is not a dominant contributor to severe COVID-19 and, more generally, that the findings of a dysregulated Ang-Tie-2 axis in bacterial sepsis may not be extrapolated to viral pneumonia. These findings also suggest that therapeutic targeting of the Ang-Tie-2 axis early in COVID-19 critical illness as has been proposed [33,34] may not be optimal. The increasing trajectory of Ang-2 concentrations from days 1 to 3 in COVID-19 patients are intriguing and may suggest that endothelial dysfunction may be a later finding in COVID-19.
This study had a number of strengths. First, the comparison of critically ill COVID-19 patients to contemporaneous critically ill patients with symptoms clinically suggestive of COVID-19 but who tested negative for SARS-CoV-2 reduces confounding due to differences in clinical management and resource allocation due to the COVID-19 pandemic. Second, all ICU blood samples were collected with uniform handling to minimize any bias due to sample handling issues. All biomarker measurements were completed using the same assay and laboratory procedures. Third, clinical outcomes were ascertained using a detailed electronic case report form by trained research coordinators and all patients had complete hospital follow up.
Limitations of this study include the small sample size which may have compromised our ability to detect minor differences in biomarker concentrations between critically ill patients with and without COVID-19. Nonetheless, we had adequate power to exclude the large effect sizes that have been observed in prior studies of inflammatory biomarkers such as IL-6 in COVID-19 and the Ang-Tie-2 axis in sepsis and ARDS. We were able to identify several significant novel associations in spite of this limitation. Another general limitation of our study as with many COVID-19 studies in the critically ill is that the limitations to imaging and lab evaluation due to the respiratory isolation mandated for those who tested positive for SARS-CoV-2 could lead to biases for outcomes such as diagnosis of VTE or initiation of RRT in the COVID-19 patients. However, this is unlikely to have affected the biomarker measurements of the main comparison between COVID-19/non-COVID-19. Another limitation is that corticosteroids or anti-inflammatory agents may have influenced plasma biomarker concentrations between COVID-19 and non-COVID-19 groups. To account for this, we adjusted for known corticosteroid use but residual confounding may still exist.

Conclusions
We present data from direct comparisons of critically ill patients with COVID-19 to critically ill patients with suspicion for COVID-19 who ultimately test negative for SARS-CoV-2 infection. We find that biomarkers of endothelial dysfunction were not characteristic of the host response in COVID-19, while biomarkers of epithelial cell injury and acute phase proteins were associated with COVID-19 critical illness. Our findings may help inform identification of prognostic and predictive biomarkers for COVID-19 and identify pathways for eventual therapeutic development.
Additional file 1. Table S1: Coefficient of variation of plasma biomarkers. Table S2: Admission Diagnoses within Cohorts. Table S3: Day 1 plasma biomarkers at study enrollment between COVID-19 and non-COVID-19. Table S4: Plasma biomarkers restricted to patients with a primary ICU diagnosis of pneumonia or respiratory failure. Table S5: Trend in plasma biomarkers over Day 1 and 3 between COVID-19 and non-COVID-19. Table S6: Association of baseline biomarkers with ARDS. Table S7: Association of baseline biomarkers with severe AKI. Figure S1: Similar set of plasma biomarkers are different between critically ill patients with COVID-19 compared to without COVID-19 adjusting for receipt of corticosteroids.