Severe COVID-19 infection mainly occurs in the elderly or in patients with vascular comorbidities including hypertension or diabetes (1). The SARS-CoV-2 virus infects type I and type II lung pneumocytes and possibly also endothelial cells that harbor the SARS-CoV-2 receptor ACE-2 (2, 3). In more severe forms of COVID-19 infection, the SARS-CoV-2 nasopharyngeal viral load is high, and the virus can also be found in peripheral blood (4). This viremia is frequent in fatal forms of infection (5).

We and others have shown that at the acute phase, the biological profile is dominated by an inflammatory response involving cytokines such as IL-6, TNF-α and IFN-γ induced chemokines such as IP-10 (6–8). Concomitantly, a global lymphopenia during the most severe forms of the disease is observed (9). Whether the lymphopenia observed reflects the pulmonary trafficking of immune cells to the lungs or to the lymph nodes, via cytokines such as IFN-I that may promote T-cell attachment to endothelium (10), is still unknown. Moreover, the immune mechanisms underlying the partial efficacy of steroids is poorly defined (11).

During the more severe forms of the disease, a strong humoral and cellular immune response to the virus is observed. Patients achieve an IgA and IgG seroconversion within 20 days after symptoms appear (12). A robust activated CD4 and CD8 T-cell response has been associated with clinical severity, while mortality could not be linked with a distinct immunophenotype (13, 14). Thus, for very severe COVID-19 infection, the immune mechanisms leading to death are still largely unknown, particularly whether the local immune response is protective or if, on the contrary, it leads to immunopathology mechanisms resulting in lung injury, as observed during fatal forms of pulmonary influenza infection, which has been linked to pulmonary immune complex deposition (15), the trapped antibodies being influenza specific (16).

Here, we characterized the systemic and pulmonary local immune response in an intensive care unit (ICU) in adults with favorable or fatal COVID-19 infection. We found evidence for a Th1 CD4 polarization of pulmonary T lymphocytes, with a high frequency of SARS-CoV-2-specific T cells at the alveolar level. Furthermore, we show a memory CD4 T-cell vasculitis in lungs from 3 additional autopsies.

We studied sequential blood samples from a total of 74 patients including 46 first and 28 second waves, respectively, with proven critical COVID 19 infection (median age 54 years, sex ratio 3.3, Figure 1A, Tables 1, 2). We also analyzed BAL from 35 patients (11 deceased) at different time points during their ICU stay. The second-wave patients received steroids for the treatment of COVID-19 infection for a median 11 days.

We first studied the blood and BAL IgA and IgG spike-specific SARS COV2 serology (n=35). All first-wave patients were seroconverted within 15 days with high serum antibody levels, the IgG response being time-shifted when compared to the IgA response (Figures 1B, C). However, deceased patients displayed a trend for a lower IgA response (p=0.052, Figure 1D) and significantly lower SARS-CoV-2-specific IgG signal-to-cut-off ratios at the first serum sampling at ICU admission than survivors (p=0.004, Figure 1E) and a trend for a lower IgA response (p=0.052, Figure 1D). The BAL retrieved from first-wave deceased patients contained SARS-CoV-2-specific IgA and IgG (Figures 1F–I). We investigated whether these low levels of BAL antibody response were linked to immune complex lung deposition, but the BAL antibody titers did not increase after immune complex dissociation in the glycine buffer (Supplementary Figure 1). In contrast, we found no difference during the second wave between the survivors and the deceased patients (Supplementary Figure 2).

Then, peripheral blood SARS-CoV-2-specific cell-mediated responses were measured in 58 patients according to PBMC availability using an ELISpot-IFNγ assay and pools of 18 amino-acid long peptides covering the nucleocapsid (NC) and/or spike(S) antigens. We detected responses against the NC and S in 26/36 (72%) first-wave patients, including deceased patients, but only 8/22 (36%) in second-wave steroid-treated patients (Figure 2A). Individually, effector responses were mainly directed against the S1 spike protein and there was no difference between groups (first/second wave, deceased/survivors, Figures 2B, C and Supplementary Figure 4). Seven non-exposed donors were tested similarly and showed 2 NC responses (Supplementary Table 2).

To further describe the local pulmonary immune responses, we first phenotyped BAL cells and found a lower proportion of BAL lymphocytes during the second wave in steroid-treated patients (p=0.012, Figure 2D). There was also a trend for lower BAL lymphocyte percentages in non-survivors (p=0.191, Supplementary Figure 3A). The CD4/CD8 ratio among BAL lymphocytes showed a large predominance of CD4 over CD8 T cells during the first wave (p=0.0003, mean ratio 1.80, Figure 2E and Supplementary Figure 3B), while similar CD4 and CD8 BAL T-cell percentages were observed in steroid-treated patients during the second wave (median CD4 43%, mean ratio 1.14, non-significant with the first wave, Figure 2F and Supplementary Figure 3B). Accordingly, the CD4/CD8 ratio elevation seems more pronounced during the first wave (Figure 2G). Importantly, when available, those alveolar T cells could be analyzed in 7 cases for their specificities and showed high frequencies of SARS-CoV-2-specific T cells, both against the spike (median 56% of the BAL T- cell response) and the nucleocapsid protein (median 44% of the BAL T cell response) and both in survivors (n=4) and non-survivors (n=3) (Figure 2H). The frequencies of those specific T cells were more than 10-fold higher in BAL lymphocytes than in PBMCs (Figure 2I), even in steroid-treated patients.

We next assessed the serum (Figure 3A) and BAL (Figure 3B) cytokine levels and found similar profiles between 12 survivors and deceased first wave patients despite some BAL/serum ratios were either below (IL-18) or above 1 (IL-8, IL-1b, Figure 3C). Noteworthy, in the 15 second-wave steroid-treated patients studied, we found significantly lower IL-6 and IFN-γ serum levels when compared to the second-wave levels (p<0.007, Figure 3D), showing the anti-inflammatory and the peripheral modulation of the Th1 cell responses of steroids during severe COVID-19 infection. On the contrary, the BAL cytokine levels were similar in steroid-treated patients (Figure 3E), concordant with the persistence of local pulmonary inflammation under treatment (Figure 3F).

To further characterize in situ the pulmonary immune environment during a fatal COVID-19 infection, we performed three autopsies of additional deceased non-steroid-treated patients (ages 58, 70, and 73). Two patients died in the ICU and one in standard hospitalization. The SARS-CoV-2 RT-PCR was positive on all three pulmonary samples, and diffuse alveolar damage lesions were observed and accompanied with vascular thrombosis in one patient. In all three cases, we observed a CD3+ T lymphocytic vascular inflammation with perivascular infiltrates, surrounding middle-sized vessels, some of them colocalizing with thrombosis (Figures 4A, B). Those cells were characterized as CD3+CD4+CD45RO+ memory T cells in imaging mass cytometry (Figure 4C). Noteworthy, these perivascular infiltrates were positive for SARS-CoV staining with the positivity of endothelial cells or perivascular cells (Figure 4D). Furthermore, several atypical aspects mimicking vascular malformations were observed in the lung parenchyma or in the subpleural tissue (Figure 4E). Those vascular structures were surrounded by the infiltrates of CD20+ cells (Figure 4E). We next retrieved archived lung sections from a case deceased from influenza H1N1 pulmonary infection in 2009. None of these perivascular lymphoid infiltrates were observed in this postmortem fragment (Figure 4).

Taken together, these results suggest a new immunopathological mechanism for pulmonary severe and fatal COVID-19 infection in an ICU. Indeed, we showed the alveolar recruitment of SARS-CoV-2-specific T cells against both the spike and the nucleocapsid. This may be due to the preferential replication of SARS-CoV-2 in alveolar pneumocytes and respiratory epithelial cells, as well as in endothelial cells as observed in lung fragments. This T-cell response was a Th1 response as assessed by the ELISpot-IFNg assay and might be predominantly of CD4 T-cell origin as suggested by the immune phenotyping. Also consistent with this hypothesis, the 18 amino-acid long peptides we used might stimulate better the CD4 T cells than the CD8 ones. The IFN-γ, IL-22, and IL-10 levels we found in BAL might also suggest the presence of some TfH and Tregs or Th2 CD4 T cells in the lungs. Altogether, these data suggest that despite severe COVID-19 pneumonia, SARS-CoV-2- specific CD4 T cells migrate to the lungs but remain detectable in the periphery. We also show that steroid treatment seems to damper systemic inflammation and specific CD4 T-cell responses, with a lower percentage of BAL lymphocytes during the second wave, even if an increase of non-lymphoid cells could not be excluded. We found no other explanations for the difference between the two waves with the same virus strain, the Wuhan original strain. The delay from first symptoms was longer during the second wave (p<0.001, Tables 1, 2), probably because of slightly different recruitment (all but one ECMO patient during the second wave). However, and taken together, our results suggest that the pathogenic role of CD4 T lymphocytes could be evoked during severe COVID-19 because steroids classically damper the T-cell responses and are associated with a better outcome.

Accordingly, during fatal forms, the strong alveolar T-cell response we describe is indistinguishable from the survivors’ one, suggesting that a lack of specific response is not responsible for the fatality. In contrast, the 3 additional autopsy cases allowed us to describe for the first time a pulmonary vasculopathy with a memory CD3+CD4+CD45RO+ T-cell lymphocytosis. This was not described by others with the same technique (Rendeiro Nature 2021, Wang Front Microbiol 2020) with, however, the description of a macrophage lung infiltration and a lack of CD8 infiltration during COVID-19 infection. This particular pattern we observed may be linked to the systemic and local virus replication, as pulmonary endothelial cells being infected in higher rates, and might be favored by endothelial injury in the context of vascular comorbidities. This angiopathy may be linked to the vascular thrombosis observed in more severe forms of COVID-19 infection (22). Furthermore, this pulmonary memory CD4 response observed in fatal forms of the disease is reminiscent of a work by Bacher et al. showing that low-avidity SARS-CoV-2-specific CD4 T cells are detectable in periphery only during more severe forms of the disease. A low avidity profile was also observed for SARS-CoV-2-specific T cells in non-exposed donors, suggesting that pre-expanded CD4 T-cell responses could accumulate during COVID-19 in the context of inflammation and be inefficient to clear the virus (23). This is hypothesis is corroborated by the SARS-CoV-2- specific T cells we and others observed in non-exposed donors. We propose that those cells could accumulate in the lungs at both the alveolar and perivascular levels during more severe forms of the disease.

We also show herein local alveolar specific immunoglobulins and confirm the precociously detectable humoral SARS-CoV-2-specific systemic response (24). A lower IgG and IgA serum dosage was observed during the first wave in deceased patients, although we could exclude immune complex lung deposition. Such a lack of immune complex deposition differs from our previous findings in lethal cases of H1N1 pneumonia (16). These differences were not observed in steroid-treated patients, suggesting a beneficial effect of steroids. Interestingly, we show here that the local humoral response may be linked with sparse perivascular B-cell formations surrounding the T-cell infiltrates, which could correspond to the tertiary lymphoid structures reported in the lungs of animal models of influenza infection (25). Supporting also this concept is the predominance of CD4 T cells in BAL and the levels of IL-22, suggesting that some TfH cells might be at play. Hence, the local specific response to the virus seems to be organized appropriately during severe COVID-19 pneumonia.

There are limitations in such a study because we could not have access to pulmonary biopsies in surviving patients or in steroid-treated patients. One can suspect that the lung T-cell angiopathy occurs in surviving ICU forms and that the hallmark of fatal forms is rather a pulmonary memory CD4 T-cell infiltrate responding to the virus. We could not determine the exact site of viral replication in this study. However, the viral staining on endothelial cells and in the perivascular area in three deceased patients allows us to propose that the observed T-cell pulmonary angiopathy could be the consequence of an in situ migration of effector/memory T cells, thus corroborating the hypothesis of a preferential pulmonary vascular virus replication in this context. Whether the T-cell perivascular migration occurs precociously in the onset of the disease and whether comorbidities such as diabetes, hypertension, or age impact this phenomenon remain to be determined. Finally, the impact of steroids on the local T-cell response may be addressed in a future randomized study.

In summary, our study proposes a new immunopathological mechanism for fatal forms of COVID-19 infection. The demonstration of high frequencies of SARS-CoV-2-specific T cells at the alveolar level shows that despite the severity, an appropriate or even pathogenic T-cell trafficking to the lungs could be at play. The novel demonstration of a memory CD4 T-cell pulmonary angiopathy colocalizing with SARS-CoV-2 in a context of active immune response opens new insights for future therapeutic interventions.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

The protocol was approved by our institution’s ethics committee (Immuno-COVID-REA, CER-Sorbonne Université, no. CER-SU-2020-31). The patients/participants provided their written informed consent to participate in this study.

AG, IP, DS, A-GM, GG, and C-EL designed the study. CS, AR, CN, IM, AG, KD, CP, and PH did the experiments. LA, OB, and PR did biobanking. AG, BC, PD, BA, A-GM, and C-EL compiled the data and discussed the results. IP, DS, and FC realized the autopsies and pathological examinations. JM, AB, GV, AL, and C-EL recruited ICU patients. AG and CEL wrote the paper. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.