Serum TGF‐β as a predictive biomarker for severe disease and fatality of COVID‐19

For targeted intervention in coronavirus disease 2019 (COVID‐19), there is a high medical need for biomarkers that predict disease progression and severity in the first days after symptom onset. This study assessed the utility of early transforming growth factor β (TGF‐β) serum levels in COVID‐19 patients to predict disease severity, fatality, and response to dexamethasone therapy. Patients with severe COVID‐19 had significantly higher TGF‐β levels (416 pg/mL) as compared to patients with mild (165 pg/mL, p < 0.0001) or moderate COVID‐19 (241 pg/mL; p < 0.0001). Receiver operating characteristics area under the curve values were 0.92 (95% confidence interval [CI] 0.85–0.99, cut‐off: 255 pg/mL) for mild versus severe COVID‐19, and 0.83 (95% CI 0.65–1.0, cut‐off: 202 pg/mL) for moderate versus severe COVID‐19. Patients who died of severe COVID‐19 had significantly higher TGF‐β levels (453 pg/mL) as compared to convalescent patients (344 pg/mL), and TGF‐β levels predicted fatality (area under the curve: 0.75, 95% CI 0.53–0.96). TGF‐β was significantly reduced in severely ill patients treated with dexamethasone (301 pg/mL) as compared to untreated patients (416 pg/mL; p < 0.05). Early TGF‐β serum levels in COVID‐19 patients predict, with high accuracy, disease severity, and fatality. In addition, TGF‐β serves as a specific biomarker to assess response to dexamethasone treatment.


Introduction
Since the declaration of the pandemic situation in March 2020 by the WHO and as of May 24, 2023, more than 6.9 million deaths in connection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection have been registered worldwide [1].A critical phase of infection is the first week after onset of symptoms [2].In this early phase of the disease, patients can rapidly progress from mild symptoms (e.g., mild fever and cough) toward life-threatening acute respiratory distress syndrome requiring hospitalization and even transfer to intensive care, ventilation, or extra corporeal membrane oxygenation.Therapeutic intervention with antivirals, therapeutic antibodies, or anti-inflammatory agents early after symptom onset is essential to prevent longterm hospitalization, death, or post-COVID-19 (coronavirus disease 2019) morbidities [3][4][5].To identify patients at risk as early as possible for targeted intervention, the identification of predictive biomarkers detectable within a few days post symptom onset is urgently needed.
We have recently described a substantial increase in serum concentrations of transforming growth factor β (TGF-β) in the first 2 weeks after the onset of symptoms in COVID-19 patients admitted to the intensive care unit (ICU).This early increase of serum TGF-β impairs the ability of NK cells to eliminate virusinfected cells [6].TGF-β also drives a chronic immune reaction, no longer directed against the virus, in COVID-19 patients requiring intensive care [7].Thus, TGF-β qualifies as a potential predictive biomarker in COVID- 19.
Here, we systematically assessed if serum TGF-β concentrations in the first 2 weeks after onset of symptoms are predictive of COVID-19 disease severity later on.Moreover, we investigated serum TGF-β concentrations for their predictive value for fatal outcomes and the response to therapy with dexamethasone.Our overall aim was to test if TGF-β serum concentrations can be used to support therapy decisions in COVID-19 and to identify patients at risk for a fatal outcome.

Study participants
Written informed consent was provided by all patients or legal representatives for participation in the study.The recruitment of study participants was approved by the Institutional Review Board of Charité (EA2/066/20, EA2/072/20, EAEA4/014/20, and EA2/092/20).Twenty-seven ambulant patients with COVID-19 (World Health Organization [WHO] disease severity 1 and 2 according to the WHO clinical ordinal scale) [8], 11 hospitalized patients (WHO 3-4) with moderate COVID-19, and 25 patients with severe COVID-19 who required ventilation (WHO 5-7) were analyzed in this study.Patients receiving dexamethasone were treated according to recommended dosing by the WHO [9].All patients with COVID-19 were tested positive for SARS-CoV-2 RNA via nasopharyngeal swabs.Fourteen patients who presented with influenza-like symptoms but tested negative for SARS-CoV-2 were analyzed as controls.

TGF-β and cytokine measurements
TGF-β was detected using a Human TGFβ1 Simplex ProcartaPlex kit (Thermo Fisher Scientific).Before measuring serum TGF-β1, the bioactive form of TGF-β1 was generated by incubating the serum with 1 N HCl for 10 min at room temperature followed by neutralization with 1.2 N NaOH.Cytokines were detected using a bead-based multiplex cytokine array (Human Cytokine 25-Plex ProcartaPlex Panel 1B; Thermo Fisher Scientific).Prior to the assay, serum samples were diluted 1:3 in the dilution buffer provided with the kit.
The samples were incubated with antibody-coated magnetic beads for 30 min at room temperature with shaking, then incubated overnight at 4°C followed by a 1 h incubation period at room temperature the next day.All subsequent incubation steps were performed according to the manufacturer's instructions.The assay plates were read using a Luminex MAGPIX system and quantified using xPONENT analysis software (Luminex).

Data analysis
GraphPad Prism 8 (GraphPad Software) was used for data analysis.Statistically significant differences between groups were calculated using the Dunn's multiple comparisons test and the Mann-Whitney U test.For all tests, p < 0.05 was considered as statistically significant.The receiver operating characteristic (ROC) analysis was applied to determine the area under the ROC curve (AUC) as well as the sensitivity and specificity at optimal TGF-β expression cut-off as defined by the maximal Youden's index [10].The Youden's index was calculated using the OptimalCutpoints R-package [11].The generalization to an independent data set was estimated using threefold cross-validation with 1000 repetitions while ensuring that the prevalence of the patient sample sets remained consistent.Own scripts in Java were used (the scripts will be shared upon request).

Results
Serum  S1).Considering the time window in which SARS-CoV-2 patients are diagnosed and show symptoms, that is, 1-2 weeks after infection [12,13], and based on our previous observation that serum TGF-β is significantly upregulated in this early phase of the infection [6], we analyzed patients during the first 2 weeks after symptom onset (Supporting Information Table S1).Using a commercially available bead-based assay (Human TGF-β1 Simplex ProcartaPlex kit; Thermo Fisher Scientific), we determined the concentrations of TGF-β in the serum.Mild COVID-19 patients (n = 27) had concentrations of 165.4 pg/mL ± SD 105.4,and patients with moderate COVID-19 (n = 11) had slightly but not significantly higher levels of 240.6 pg/mL ± SD 150.2.In contrast, patients with severe COVID-19 (n = 25) had markedly and significantly higher TGF-β serum levels of 416.3 pg/mL ± SD 161.7 (p < 0.0001), as compared to patients with mild or moderate COVID-19.This significant difference in predictive serum TGF-β concentrations was robust for the first 2 weeks after the onset of symptoms.We did not observe significant differences of TGFβ levels in COVID-19 patients compared to patients with flu-like symptoms but not infected with SARS-CoV-2 (n = 14) who had even lower concentrations of serum TGF-β, that is, 105.9 pg/mL ± SD 76.44 (Fig. 1).Taken together, high serum TGF-β levels, during the first 2 weeks after the onset of symptoms, predicted progression to severe COVID-19 requiring ICU care.
A ROC curve analysis was performed to assess the prognostic accuracy of serum TGF-β concentrations across different patient groups.In addition, the positive predictive value and negative predictive value (PPV and NPV) and cut-off values were calculated (Table 1).The area under the curve (AUC) for the cut-off value of 254.75 pg/mL (Fig. 2a, Table 1) was 0.92 (95% confidence interval [CI] 0.85-0.99,PPV 0.84, NPV 0.85) for mild versus severe (i.e., requiring ICU treatment) COVID-19.The AUC for 202.11 pg/mL as a cut-off value was 0.83 (95% CI 0.65-1.0,PPV 0.85, NPV 0.78) for moderate versus severe COVID-19.All AUC values indicated a high accuracy of prediction endorsed by the highest Youden's index [10].The identified cut-off values showed strong predictive accuracy of serum TGF-β validated by the high PPV and NPV values (Table 1).Of note, this refers to TGF-β levels in hospitalized patients before the onset of dexamethasone treatment.TGF-β-mediated pathology (e.g., airway remodeling and organ fibrosis) is known to be reduced by this treatment [14,15].Notably, serum levels of other cytokines elevated during the first 2 weeks of COVID-19, that is, IFN-α, IFN-y, IL-6, IL15, and IL-18, did    S2 and Fig. S1) [6].
We further validated serum TGF-β as a biomarker for predicting severe COVID-19, that is, requiring intensive care, by performing multivariate logistic regression with a large virtual independent data set [16].We kept the prevalence of the virtual sample sets at the same level as the real samples and used threefold cross-validation with 1000 repetitions.The predictive power of the validating data set was comparable to that of the original data set: AUC of 0.89 (95% CI 0.88-0.89,cut-off: 271.3 pg/mL, PPV 0.88, NPV 0.76) for mild versus patients requiring ICU, and 0.75 (95% CI 0.75-0.76,cut-off: 287.5 pg/mL, PPV 0.84, NPV 0.51) for hospitalized patients with moderate disease course versus those requiring ICU care later on (Fig. 2b, Table 2).

TGF-β as biomarker to assess the response of COVID-19 to treatment with dexamethasone
Dexamethasone is one of the approved early therapeutic interventions for COVID-19 [17].Dexamethasone has well-established anti-inflammatory effects, whereas less is known about its effect on TGF-β expression.Analyses of serum TGF-β concentrations in dexamethasone treated (n = 18) versus untreated (n = 25) COVID-19 ICU patients revealed a profound suppressive effect of dexamethasone on early serum TGF-β levels, with 301.2 pg/mL ± SD 162.4,as compared to 416.3 pg/mL ± SD 161.7 in untreated patients (Fig. 4).The mean TGF-β serum level of dexamethasonetreated patients dropped below the cut-off value for predicted mortality (454.89pg/mL).Thus, early serum TGF-β levels also allow to assess the response of COVID-19 patients to treatment with dexamethasone, and the suppression of TGF-β expression may well add to the therapeutic benefit of dexamethasone.1), and the dashed line indicates the threshold for mortality (454.89pg/mL, Table 1).The p-value was calculated using the Mann-Whitney U test.

Discussion
Our study positions TGF-β as a unique and specific biomarker that predicts, early during infection, the course and severity in patients with COVID-19.High TGF-β serum levels during the early days of COVID-19 forecast severe disease including the need for intensive care and the risk of death.The use of TGF-β testing in early COVID-19 may help to guide treatment decisions and identify patients at risk of severe disease with a high need for therapeutic intervention.
SARS coronaviruses including SARS-CoV and SARS-CoV-2 have caused epidemic and pandemic situations with considerable impact on global health and welfare [18].Both SARS-CoV and SARS-CoV-2 have developed an efficient mechanism to evade the immune response of infected individuals by inducing expression of the immunosuppressive cytokine TGF-β in infected cells.They use several mechanisms to induce expression of TGF-β and signal transduction in infected cells, such as integrin-mediated induction by the spike protein [19], nucleocapsid protein pairing with Smad3 [20,21], induction of TGF-β target genes (including TGF-β itself) [6], and induction by orf3 and orf8 encoded proteins [22].
TGF-β itself is a major and pleiotropic player in the regulation of immune responses, in particular in mucosal tissues.It targets antibody class switch recombination to IgA [23], inhibits T cell [24], and NK cell activities [6], and supports the differentiation of regulatory T cells [25].Apart from its central role in regulating immune responses, it is involved in tissue regeneration.It may thus also regulate fibrosis and thrombosis in COVID-19 [26], and the reactivation of herpes viruses including Epstein-Barr Virus [27], all hallmarks of COVID-19 pathogenesis.TGFβ can be expressed by many cell types, but information on signals inducing TGF-β is scarce.Apart from SARS-CoV, particular bacteria have been described to induce TGF-β [28].A potent autocrine inducer of active TGF-β is TGF-β itself [29,30], by binding to its receptor, inducing dimerization of Smad2 and Smad3 and transcription of the TGF-β target genes including TGFB1 [31].We previously demonstrated that "infectious" TGF-β expression originates from SARS-CoV-2 infected cells in the lungs of severely affected COVID-19 patients [6].The direct induction by the pathogen and self-sustained amplification of TGF-β expression in COVID-19, and the functional implication of TGF-β in the pathogenesis of the disease, make it an ideal candidate to predict at an early-stage progression of the disease and response to therapy.
Early TGF-β expression has severe pathogenic consequences, inhibiting the killing of infected cells by NK cells [6] and redirecting the immune response in severely affected patients toward a chronic autoimmune response [20,32].It may also impact fibrinogenesis [22], vascular leakage [19], and thrombosis in COVID-19 [26].Therefore, detection of TGF-β levels at an early stage after the onset of symptoms (i.e.1-2 weeks) would allow risk stratification as well as early and individualized treatment to prevent severe disease and death.
For pandemic preparedness against current and future emerging SARS coronaviruses, antiviral agents and vaccines are needed, as are biomarkers that predict disease progression and response to therapy in infected individuals.Here, we show that serum TGF-β levels in infected individuals 1-2 weeks after symptom onset are an excellent predictive marker of disease severity (cutoff: 254.75 pg/mL, AUC: 0.92, PPV: 0.84 NPV: 0.85), mortality (cut-off: 454.89 pg/mL, AUC: 0.75, PPV: 0.83 NPV: 0.71), and response to steroid treatment.TGF-β is the first biomarker offering this predictive power in the early phase of the disease.Proinflammatory markers of disease progression, that is, CRP and IL-6, have been described earlier [33][34][35], but since these markers indicate inflammation as such, they are less specific for SARS coronaviruses [36,37].
To be successfully used in clinical routine, the measurement of a biomarker should ideally be quick and easy to perform.ELISA measurements are routinely performed in most diagnostic laboratories, and TGF-β ELISA kits are available from a variety of manufacturers worldwide.Alternatively, TGFbeta assays based on Luminex, Mesoscale, or Simoa technology are also available.In this study, TGF-β levels were analyzed using a bead ELISA based on Luminex technology.The assay results in this study were obtained within 24 h of overnight incubation of the samples with the antibody-coated beads.However, according to the manufacturer's protocol, it is also possible to perform all experimental steps within the same day and thus obtain rapid results without loss of sensitivity, which is important in view of the need for timely decision-making on therapeutic measures.
Limitations of the present study include the relatively small size of the cohort, and the lack of an independent prospective cohort, due to routine usage of steroids since 2021 at the University Hospital Charité.The predictive power of TGF-β serum concentrations thus had to be validated by cross-validation.Although serum TGF-β concentrations were determined using one assay system, the signals have been shown to be comparable to those obtained with other assay systems [38,39].In addition, a similar study using a commercially available ELISA kit to measure serum TGF-β levels also proved useful in predicting the severity of COVID-19 [40].
In conclusion, we observed that patients showing a TGF-β serum level of above 254.75pg/mL will progress with a 92% chance (predictability) into severe disease within the first 2 weeks after onset of symptoms.This will require intensive clinical care and even may lead to death.TGF-β monitoring within the first 2 weeks post symptom onset could guide early intervention with dexamethasone, antivirals, and therapeutic antibodies.This may be particularly relevant for epidemic and pandemic challenges caused by new SARS-coronaviruses before efficient vaccines become available.However, further studies are needed to confirm that TGF-β specifically indicates the severity of COVID-19 and not of other viral infections.Last but not least, the excellent correlation of TGF-β levels and activities with disease activity in COVID-19 suggests targeting of TGF-β to treat COVID-19 [41].

Figure 1 .
Figure 1.Serum TGF-β is significantly increased in COVID-19 patients hospitalized in the ICU.Mean serum TGF-β levels of patients measured by a ProcartaPlex bead-based ELISA within 2 weeks post symptom onset are shown for influenza-like patients (flu-like, n = 14) not infected with SARS-CoV-2, as well as for COVID-19 patients with mild disease (ambulant, n = 27), and hospitalized patients with moderate (n = 11) or severe disease (ICU, n = 25).The dotted line indicates the calculated TGF-β cutoff for ICU care (≥254.75pg/mL, Table1).Statistics were performed using the Dunn's multiple comparisons test.

Figure 2 .
Figure 2. Serum TGF-β is highly predictive for ICU requirement in COVID-19 patients.(A) Receiver operating characteristic (ROC) analysis of TGF-β within 2 weeks post symptom onset for COVID-19 patients with mild disease (ambulant) versus patients on the ICU (left) and for hospitalized patients with moderate disease versus patients with severe disease requiring ICU care (right).(B) Validation of patient data using a virtual data set based on real patient data.Area under the curve (AUC), confidence intervals, as well as sensitivity and specificity values are displayed in the graph.

Figure 3 .
Figure 3. Serum TGF-β as predictive biomarker for fatal COVID-19 outcome.(A) Mean serum TGF-β levels measured a ProcartaPlex bead-based ELISA within 2 weeks post symptom onset of convalescent patients (n = 13) and deceased (n = 10) patients.The dotted line indicates the calculated TGF-β cut-off for ICU care (≥254.75pg/mL, Table 1).The p-value was calculated using the Mann-Whitney U test.(B) ROC analysis of TGF-β of convalescent and deceased COVID-19 patients within 2 weeks post symptom onset.Area under the curve (AUC), and confidence intervals, as well as sensitivity and specificity values are displayed in the graph of the patient cohort (left) and the virtual validation cohort (right).

Figure 4 .
Figure 4. Serum TGF-β as biomarker for treatment response to dexamethasone in COVID-19 patients.Mean serum TGF-β levels of untreated (n = 25) versus dexamethasone-treated (n = 18) patients in ICU measured by a ProcartaPlex bead-based ELISA within 2 weeks post symptom onset.The dotted line indicates the calculated TGF-β threshold for ICU care (≥254.75pg/mL, Table1), and the dashed line indicates the threshold for mortality (454.89pg/mL, Table1).The p-value was calculated using the Mann-Whitney U test.

Table 1 .
Validation results of diagnostic accuracy of the COVID-19 patient cohorts.

Table 2 .
Cross-validation of diagnostic accuracy.