Histomorphological patterns of regional lymph nodes in COVID-19 lungs

Severe lymphopenia was one of the first observations among COVID-19 patients in early clinical case series [16]; its severity was shown to be the most important predictive factor of disease progression in a meta-analysis of 32 studies involving over 10,000 patients [26]. Pathophysiologically, lymphopenia could be the outcome of chemotactic dysregulation of CCL8, CCL20, and CXCL10, which were shown to be upregulated in COVID-19 lung tissue [1]. These molecules are responsible for chemotaxis and migration of (T) lymphocytes, possibly leading to extensive leukocyte trafficking into the lungs. Furthermore, lymphocytes express angiotensin-converting enzyme 2 (ACE2) [46], rendering possible direct cytopathic effects a further plausible explanation of lymphopenia.

IFN-mediated signal transduction plays a central role in the primary immune response to viral infections. Preliminary results of single-cell sequencing of peripheral blood mononuclear cells of COVID-19 patients revealed a phenotypic reconfiguration of immune cells with heterogeneous IFN-stimulated gene (ISG) signatures and downregulation of HLA class II genes, which was also associated with an increased plasmablast population [45]. Furthermore, ISG expression also appeared to be related to disease progression and viral load [34]. A subgroup of patients with early lethal disease showed high expression of ISG and cytokines, associated with high viral load and mild lung injury (ISG_high) [30]. In contrast, another subgroup with prolonged disease showed decreased expression of ISGs (ISG_low) in association with severe diffuse alveolar damage, lower viral load, and increased infiltration of CD8⁺ cells and (M2-polarized) macrophages [34].

Gene expression analyses of COVID-19 lungs revealed downregulated expression of key costimulatory molecules of antigen presentation, such as TLR7, TLR9, and CD86 [1]. CD86 is expressed by antigen-presenting cells and binds to CD28 or CTLA4 expressed by T cells. Interaction between CD86 and CD28 subsequently stimulates a T cell response [39]. Other TLRs (TLR1, TLR4, TLR5) were also found to be downregulated in COVID-19. While TLR1 binds to peptidoglycans of Gram-positive bacteria, TLR4 interacts with lipopolysaccharides (LPS) of Gram-negative bacteria and is central for recognition of the SARS-CoV‑2 spike antigen [36]. Activation of TLR5 by flagellin stimulates the innate immune response against certain bacterial species [39]. This may explain why COVID-19 patients experience increased susceptibility towards bacterial superimposition.

Furthermore, the development of hemophagocytic lymphohistiocytosis (HLH) has been observed in COVID-19 patients [29, 38]. HLH typically is characterized by an uncontrolled macrophage activation syndrome, linked to impaired antigen presentation/recognition and excessive cytokine release (cytokine storm), but—to a limited and controlled extent—is known to be a typical reaction of the innate immune defense as a so-called pattern-recognition receptor (PRR)-associated response to pathogen-associated molecular patterns (PAMPs), specifically viral RNA [4, 30, 44]. HLH can be triggered by herpesviruses, especially EBV, but also by SARS-CoV‑1 and influenza viruses such as H5N1 and H1N1 [6, 15]. In accordance with this, COVID-19 tissue studies show increased expression of CCL2 and CCL7 [1], known to be chemotactic for macrophages, thus stressing the central role of the cytokine storm in COVID-19 pathogenesis and highlighting a potential therapeutic intervention in severe disease [25]. Furthermore, an upregulation of genes encoding complement activation and phagocytosis is observed in severe disease [48]. In this context, pediatric COVID-19 patients have been observed to develop multisystemic hyperinflammation, which phenotypically resembles Kawasaki syndrome [37]. In such cases, severe vascular and cardiac manifestation was observed, occasionally together with macrophage activation syndrome.

In one of the first autopsy studies, an absence of germinal centers, an increased presence of plasmablasts, and intranodal capillary stasis were described in hilar lymph nodes and splenic parenchyma [30]. These histomorphological changes may be explained by a dysregulation of BCL6⁺ follicular T helper cells [21]which play an essential role  in germinal center functionality. Additionally, early blockade of T helper cell differentiation, a predominance of T‑bet⁺ T helper 1 cells, and an extrafollicular accumulation of TNFα could be shown [21], corresponding to a loss of follicular B cells by flow cytometric analyses of peripheral blood from severely ill COVID-19 patients [21]. Interestingly, TLR4 and TLR5, which are both downregulated in COVID-19 as mentioned earlier, are also essential for the germinal center response as they activate the NF-κB signaling pathway via MYD88 [13].

The previously mentioned proliferation of plasmablasts in hilar lymph nodes in lethal COVID-19 infection could represent a morphological correlate of a dysregulation of immunoglobulin (Ig) class switching. This is supported by a markedly increased plasmablast population in patients with severe disease progression, whereas a robust adaptive immune response with clonally expanded CD8⁺ effector- or emory cells, at best, is observed in mild disease [23, 49]. Gene expression analyses of COVID-19 autopsy tissue also demonstrated downregulation of CD40LG [1], which is an essential link of communication between T and B cells and significantly influences B cell maturation [39]. A defect in CD40LG results in an absence of Ig class switching, which may favor preferential extrafollicular proliferation of B cells. As a reflection of the aforementioned two features, in-depth immunological analyses demonstrated a negative correlation between the number of memory B cells and COVID-19 symptom duration [33]; the number of these cells correlated with IgG1 and IgM against the SARS-CoV‑2 spike protein. This was also reflected in antibody titer measurements [32] and flow cytometric analyses, which showed oligoclonal plasmablast expansion (>30% of circulating B cells) [22] in cases with severe disease. This, along with other immunologic signatures that correlated with disease progression, allowed a biostatistical classification of three COVID-19 immune phenotypes with different risk profiles [28].

As a link to immunopathology, an incomplete humoral immune response with low-affinity, non-(sufficiently)-neutralizing, low-titer antibodies can lead to antibody-dependent enhancement. This is defined as the generation of suboptimal antibodies, which enable virus penetration into Fc/complement receptor-bearing monocytes, macrophages, and granulocytes [19]. Indeed, data indicate an interaction of anti-spike protein antibodies and macrophages that contributes significantly to lung injury in SARS-CoV‑1 [24]. Finally, spike protein reactivity was also shown to be present in over one third of SARS-CoV-2-naïve patients. This implies the presence of cross-reactive T cells, developed in the course of immunization against other coronaviruses and may explain the more robust immune response in some patient groups [4].

Hilar, mediastinal, and cervical lymph nodes were acquired from autopsies of COVID-19 patients (n = 20). Staining protocols of histochemical and immunohistochemical assays (H&E, IgG, IgM, CD3, CD11c, CD20, CD79a, CD68, CD163, CD206, HLA-DR, SARS-CoV-2‑N antigen [Rockland 200-401-A50 rabbit polyclonal antibody, dilution 1:2000]) were performed in accordance with accredited standard operating procedure (SOP) protocols of the Institute of Pathology in Basel. Histologic characteristics (number of plasmablasts, plasma cells, edema/capillary stasis, presence of HLH and germinal centers) were evaluated using ordinal scales (0–3; Table 1).

All lymph nodes underwent RT-PCR examination to determine viral load. The protocol was described in our previous study [30].

RNA was extracted by Forma HTG (HTG Molecular Diagnostics, Tucson, USA) from 10 µm thick untreated formalin-fixed paraffin-embedded (FFPE) blocks according to established protocols and processed and analyzed as described using the HTG EdgeSeq autoimmune assay [31].

Statistical analyses were performed using SPSS version 25 (IBM, Armonk, NY, USA). All correlation analyses were performed using Spearman rho (ρ).

In 14 out of 20 cases, qRT-PCR detected SARS-CoV‑2 RNA. Interestingly, in cases without detectable SARS-CoV‑2 RNA, hospitalization time was longer (6 vs. 11 days), suggesting a biphasic disease course with progressive viral elimination, similar to previously performed analyses of the same autopsy cohort ([34]; Fig. 1b). This is supported by a negative correlation of pulmonary viral load with hospitalization time (ρ = −0.776; p < 0.0001) and a positive correlation between pulmonary and lymph node viral load (ρ = 0.514; p = 0.035). Blood counts revealed a positive correlation between hospitalization time and absolute lymphocyte count (ρ = 0.582; p = 0.014) and a negative correlation with the absolute leucocyte count (ρ = −0.577; p = 0.015).

Major histologic lymph node features included moderate-to-severe edema and capillary stasis (Fig. 2a) in all cases, explained by acute right heart failure due to severe pulmonary disease. Further findings include a proliferation of extrafollicular B blasts, IgG- and IgM-positive plasmablasts in particular (Fig. 2b,c; 12/20), consistent with rapid or primarily transient B cell immune response bypassing the germinal center response as described above [5], with largely absent or hypoplastic/hypotrophic germinal centers or secondary follicles (12/20), including decreased follicular dendritic cells and follicular T helper cells. Correlation analyses showed a negative association between the presence of secondary follicles and viral load in the lungs (ρ = −0.645; p = 0.005) and CRP (secondary follicles = −0.522; p = 0.032), which, in conjunction with previous analyses of the same autopsy cohort [34], additionally underscores a delayed germinal center response. Furthermore, a discrete-to-moderate degree of plasmacytosis and a predominance of CD8⁺ T cells was observed, accompanied by numerous HLA-DR-, CD163-, and CD206-positive M2-polarized macrophages [24], CD11c- and CD68-positive histiocytes, and the presence of hemophagocytosis in the sinuses in 12/20 cases (Fig. 2d), consistent with the aforementioned phenomena of macrophage activation in COVID-19 [29, 38]; however, the full clinical picture of HLH presented itself in only one of the 20 patients examined here [47].

An increased number of sinus histiocytes with corresponding positive immunohistochemical detection of the nucleocapsid antigen of SARS-CoV‑2 in an elderly patient with a 30-day disease course may be interpreted as antibody-dependent enhancement, as further supported by observations from Martines et al. ([27]; Fig. 3a). Although the majority of lymph nodes examined illustrated imperceptible virus antigen levels as detected by immunohistochemistry, the lungs of the same patient showed a comparatively higher viral copy number than others (990 viral genome copies/10⁶ human RNAseP copies in the lymph nodes and approximately 125000 in the lungs).

In line with the histomorphological patterns described above, gene expression profiles (Fig. 3b) showed increased expression of the following genes: STAT1 (central transcription factor in macrophage activation), CD163 (hemoglobin–haptoglobin complex receptor and marker of M2 macrophage polarization [24]), granzyme B (GZMB) but not perforin (consistent with the imbalance of both proteins in hemophagocytic lymphohistiocytosis [40]), CXCL9 and PAK1 (a chemokine and an enzyme important for the migration of cytotoxic T cells), CD8A (in line with the upregulation of PAK1), and, finally, MZB1 (marginal zone B and B1 cell-specific protein, which contributes to the composition and secretion of IgM and thus corresponds with the increased plasmablasts).