Different levels of viral entry proteins neuropilin 1 and 2 in fatal COVID-19 cases

Pulmonary tissue from fatal COVID-19 autopsies revealed alveolar damage in fulminant cases, as well as early-phase organizing chronic pneumonia in subacute cases. Histopathological features included prominent intra-alveolar edema, microthrombi, hyaline membranes, intra-alveolar fibrin deposits, and mixed inflammatory infiltrates. The myocardium demonstrated scattered mixed inflammatory cells alongside either normal or hypertrophic myocardial fibers. Similar histomorphological findings have been described previously (27, 31, 32).

To investigate NRP1 and NRP2 expression patterns in situ, standard IHC was performed on COVID-19 autopsy tissues, as summarized in Table 1. Neuropilins were expressed on epithelial, mesenchymal stromal, and immune cells in all organs examined (Table 1) with some notable differences: NRP1 was strongly expressed in macrophages and Kupffer cells, splenic white pulp and lymph node follicular lymphocytes, capillary and liver sinusoidal ECs, as well as in proximal tubules and glomeruli (Fig. S5). In contrast, NRP2 expression was more restricted and was moderately detected in intestinal smooth muscle cells and in macrophages within the lung, trachea, spleen, lymph nodes, and adrenal gland (Fig. S6).

Myocardial small capillaries, arterioles, and postcapillary venous vessels showed strong NRP1 expression by IHC in COVID-19 (Fig. 1B), whereas NRP2 was not detected (Fig. S6B). NRP1 localization on vascular EC was confirmed by CD31 co-expression using CODEX multiplex imaging (Fig. 1A). scRNAseq analysis demonstrated abundant NRP1 transcripts in vascular EC and pericytes, with NRP2 mainly restricted to lymphatic EC (Fig. 1H through K). Although NRP1 mRNA was expressed in cardiomyocytes, neither NRP1 nor NRP2 protein was detected by IHC or CODEX. In the myocardium, NRP1-positive mononuclear cells along larger vascular beds (Fig. 1D) corresponded to CD68-positive macrophages in serial sections (Fig. 1E). scRNAseq analysis confirmed strong expression of NRP1 and NRP2—but not ACE2—in cardiac macrophages (Fig. 1G through K). ACE2 transcripts were largely confined to cardiomyocytes and pericytes, consistent with prior findings (33), and barely detectable in vascular EC (Fig. 1G). Compared with influenza pneumonitis and non-infected control myocardium, NRP1-positive capillaries were more frequent in COVID-19 (F (2, 23) = 0.91, P = 0.417); Fig. S7).

NRP1 and NRP2 proteins were heterogeneously expressed in larger vessel EC of the trachea and lung of COVID-19 patients (Fig. 2B inset; Fig. S5I and S6I), while absent from small capillaries. Their transcripts were detected in vascular EC (Fig. 2L and M). Intense NRP1 expression in alveolar macrophages was not specific to COVID-19 but was also noted in other causes of death (n = 28, Fig. S2). However, we identified COVID-19-associated syncytial macrophages with high phagocytic activity in the lung that were positive for NRP1 (Fig. 2A inset, 2D) and CD68 (Fig. 2A inset, 2E) by CODEX and IHC. NRP1/CD68-positive macrophages, predominantly interstitial, expressed the microglial and macrophage marker ionized calcium-binding adapter molecule 1 (IBA1) in their cytoplasm, with ruffled membrane appearance (Fig. 2F through H), suggestive of migratory behavior (34).

NRP2 protein expression was also detected in alveolar macrophages (Fig. 2C; Fig. S6A), consistent with scRNA-seq analysis (Fig. 2M and P). NRP2 transcripts were mainly expressed in lymphatic ECs (Fig. 2P), while NRP1 mRNA was predominantly expressed by vascular ECs, fibroblasts, and myofibroblasts (Fig. 2P). Alveolar type II (AT2) and goblet cells also expressed NRP1 mRNA (Fig. 2P), although NRP1 protein was largely absent in lung epithelial cells (Fig. S5A). Focal, weak NRP2 protein expression in epithelial cells, detected by IHC, was confirmed by transcript data (Fig. 2P).

ACE2 and TMPRSS2 mRNA were largely restricted to lung epithelial cells (Fig. 2K, O and P), while ECs, stromal cells, as well as lymphocytes and macrophage populations lacked expression. No overlap was observed between SARS-CoV-2-infected cell types (Fig. 2N) and those co-expressing ACE2 and TMPRSS2 (Fig. 2O). Only low levels of SARS-CoV-2 RNA were detected in lung epithelial cells from fatal COVID-19 patients (Fig. 2N).

We detected a significant upregulation of NRP1 transcripts in vascular ECs in COVID-19 lungs compared with noninfectious controls (P < 7.5E-6, Fig. S8A), while ACE2 and TMPRSS2 expression remained unchanged. NRP1 and the vascular endothelial marker FLT1/VEGFR1 (fms-related receptor tyrosine kinase 1/VEGF receptor 1) transcript expression predominated in vascular ECs, while lymphatic ECs showed high NRP2 expression along with the lymphatic markers KDR/VEGFR2 (kinase insert domain receptor/VEGF receptor 2) and FLT4/VEGFR3 (fms related receptor tyrosine kinase 4/VEGF receptor 3) (Fig. 2P). The activating ligand for VEGFRs, VEGFA, was produced and upregulated (P < 7.5E-6) by lung epithelial cells (Fig. 2P).

Both vascular and lymphatic ECs expressed high levels of the transforming growth factor beta receptor 2 (TGFBR2) (Fig. 2P), suggesting endothelial repair in response to viral infection (35). TGFBR2 was significantly upregulated in vascular ECs (P < 7.5E-6, Fig. S8A) in fatal COVID-19, while upregulation of the neuropilin-binding ligand TGFB1 (20) was observed in monocytes (Fig. S8B). Monocytes also showed upregulation of the TGFB1 inducing factor C-C motif chemokine ligand 2 (CCL2) (P < 7.5E-6, Fig. S8B) (36).

Similar to the lung, macrophages and sinus histiocytes in the lymph nodes of COVID-19 autopsies were strongly positive for NRP1 (Fig. S5F) and NRP2 (Fig. S6F). Follicular lymphocytes in lymph nodes and spleen were NRP1-positive and NRP2-negative (Fig. 3B and C; Fig. S5E). NRP1 expression in lymphocytes was not specific to COVID-19 but was also observed in influenza and noninfectious causes of death. CODEX multiplex analysis revealed that a subset of CD20-positive splenic B lymphocytes co-expressed NRP1 (Fig. 3A), while CD8- and CD4-positive lymphocytes were negative for NRP1 (Fig. 3E). Lymphocytes in the spleen and lymph nodes showed antisense SARS-CoV-2 RNA signals, as demonstrated by RNAscope FISH (Fig. 3G).

Severity of COVID-19 inversely correlates with the plasmacytoid dendritic cell (pDC) response (37). Tissue DCs in autopsy samples, identified by CD123 expression by CODEX (Fig. 3F), were positive for NRP1 in lymph nodes (Fig. 3E) but were NRP2-negative. NRP1 expression was also detected in tracheal DCs by IHC (Fig. S5I). Transcript analysis revealed NRP1 but not NRP2 expression in pDCs (Fig. 2P). Although NRP1 transcripts were present in mast cells (Fig. 2L and P), protein expression was not detected in a subset of mast cells that instead exhibited strong NRP2 expression by CODEX (Fig. 4E through G).

To explore the role of soluble S1 and S1′ in neuropilin binding (Fig. S1), NRP2-negative HEK293 cells, which endogenously express NRP1, were transfected to express NRP2. These cells express low to undetectable levels of the SARS-CoV-2 (co-)receptors ACE2, FURIN, and TMPRSS2 (Fig. S3). HEK293 cells not expressing NRP2 (Fig. S4I) served as internal negative controls to quantify NRP2 binding to spike fragments. As previously shown (13, 15), endogenous NRP1 on HEK293 cells bound the soluble S1 spike fragment (Fig. 5A through C). However, S1′ was not bound by NRP1 (Fig. 5E through G). Only recombinant expression of NRP2 enabled HEK293 cells to additionally bind S1′ soluble spike fragment (Fig. 5N through P), with preserved S1 binding (Fig. 5I through L). Manual cell counts revealed NRP2 expression and S1′ binding in a similar fraction of NRP2-transfected cells (5%; 4%).

The cohorts in this study, comprising COVID-19 patients (n = 20) and influenza/noninfectious controls (n = 13), were relatively small. NRP2 expression detected by IHC appeared low across samples, with variable intensity likely due to differences in post-mortem intervals. Such variability may introduce noise into the data analysis and obscure biological differences. Further preclinical studies are necessary to evaluate NRP2 modulation as a potential therapeutic approach.

Combining mRNA and protein expression data in situ, we identified NRP1 and NRP2 on cell types exhibiting minimal ACE2 and TMPRSS2 expression in the lung, heart, and hematolymphoid organs. In summary, NRP1 was detected in myeloid cells, B lymphocytes, and vascular ECs, while NRP2 expression was largely restricted to macrophages and mast cells. Our study suggests that primary infection with SARS-CoV-2 in the upper respiratory tract depends on ACE2, while neuropilins are important for systemic viral dissemination, monocyte/macrophage-induced vascular damage, microangiopathy, and thrombosis. In this context, neuropilins might play a multifaceted role in SARS-CoV-2 pathogenesis, contributing to both acute and chronic complications. Neuropilins, particularly NRP2, represent promising druggable targets in the treatment of pulmonary fibrosis. As macrophages and mast cells are critically involved in viral spreading, excessive cytokine storm, and fibrotic remodeling of infected tissue, modulation of these cells by targeting NRP2 might offer a new therapeutic strategy in acute COVID-19 as well as in PASC. Further research is necessary to fully understand the functional roles of NRP1 and NRP2 in disease progression.

Autopsy tissues were analyzed using tissue microarrays (TMA) from three different institutions in Germany: University Hospital Bonn (UKB), University Hospital Aachen, and Hannover Medical School (MHH). A total of 20 patients who died from COVID-19 in 2020 were compared with seven patients who succumbed to influenza type H1N1, seasonal A or B. Additionally, six cases of noninfectious deaths were used as controls. The COVID-19 cohort included 11 females and 9 males (mean age: 70.25 ± 11.22 years). The influenza cohort comprised five females and two males (mean age: 50.86 ± 15.24 years), while the uninfected cohort included two females and four males (mean age: 71.25 ± 13.91 years). Infection with SARS-CoV-2 or influenza was confirmed by nasopharyngeal swabs (RT-PCR for SARS-CoV-2 or influenza RNA, respectively). Patient characteristics are provided in. Table S1

TMAs of formalin-fixed paraffin-embedded organ samples were assembled using tissue cores 3–4 mm in diameter. Standard paraffin sections (3–4 µm) following standard embedding (fixation: 4% buffered formalin) were used. Morphology was evaluated by hematoxylin-eosin staining. For IHC staining of CD34 and CD68, slides were processed as previously described (66). Serial section staining for CD34 (QBEnd10, Agilent, Santa Clara, USA; 1:200) and CD68 (PG-M1, Agilent; 1:100) was performed using the semiautomatic Autostainer 480S (Medac, Wedel, Germany; for staining conditions see Table S2). Serial section staining for NRP1 (EPR3113, Abcam, Cambridge, UK; 1:200) was performed after deparaffinization with EZ prep (Roche, Basel, Switzerland) and treatment with CC1 buffer (pH 8, Roche) using the automatic BenchMark Ultra staining platform with the OptiView detection kit (Roche; Table S2). TMAs from COVID-19 patients 1–3 (Table S1A) were stained manually in the case of NRP2 with 15 µg/mL (aNRP2-36v2, aTyr Pharma; for protocol see reference 67). Photomicrographs were acquired using either a BX51 microscope (Olympus, Hamburg, Germany) with Zeiss AxioCam MRc5 and Axiovision software (Carl Zeiss, Oberkochen, Germany) or digitally scanned on a Leica scanner (Aperio GT 450 DX, Leica Biosystems, Wetzlar, Germany) and analyzed with QuPath v0.4.3 (68).

For quantification of NRP1-positive myocardial capillaries (Fig. S7), three representative squares in each tissue core were selected in QuPath (each 0.0625 mm²) and positive capillaries were counted manually. A mean of three squares was calculated per patient. If multiple cores per patient were available, means were averaged across cores. A total of 35 myocardium samples were analyzed (COVID-19 n = 13 patients, influenza n = 7 patients, noninfectious controls n = 6 patients), and an analysis of variance (ANOVA) was conducted. Insufficient tissue samples were excluded from analysis and staining results confirmed by two independent observers (A.D., I.G.). Variable interpatient NRP1 intensity of staining was observed, possibly reflecting differences in autolytic degradation of autopsy tissue.

Semiquantitative scoring of NRP1 intensity in alveolar macrophages in lungs was conducted in a blinded, independent, and randomized manner. Representative peripheral lung areas per core were photographed in QuPath and evaluated blindly. A total of 56 lung samples were analyzed (COVID-19 n = 16 patients, influenza n = 7 patients, noninfectious controls n = 5 patients). One noninfectious control case (no. 3) was excluded due to lymphangiosis carcinomatosa in the lung. Staining intensities were scored twice independently as follows: 0, negative; 1, low; 2, moderate; and 3, strong (Fig. S2). Discrepant core results were reviewed for a final consensus score.

1–3 µm sectioned FFPE TMAs of COVID-19 patients 2 and 3 (Table S1A) were prepared and stained for CODEX-enabled multiplexed tissue imaging following manufacturer's instructions. The full protocol can be found in (69). A list of antibodies is provided in Table S3. Nuclei were detected using DAPI. Images were analyzed using the Enable Medicine platform (Menlo Park, CA, USA).

We re-analyzed a published and annotated scRNAseq data set (30) using preprocessed Python Scanpy objects downloaded from the Single Cell Portal. Data included 19 tissue samples from 18 heart donors (12 males and 6 females) and 24 tissue samples from 16 lung donors with COVID-19 (11 males and 5 females). Clinical metadata and a detailed protocol of tissue collection, tissue procession, RNA extraction, and computational methods for preprocessing, quality control, doublet identification, and dimension reduction are provided in Delorey et al. (2021) (30). Data were normalized using the Scanpy sc.pp.normalize_total function. Annotated doublets and cell clusters containing less than 10 cells were removed. For comparison of COVID-19 lung tissue with healthy tissue samples, Excel files containing log2 fold changes from a single cell-based differential expression model were analyzed (30). All analyzes were performed in Python v3.11.6 (70) with the Scanpy package v1.9.8 (71). The python scripts provided by the original authors on GitHub were modified for re-analysis with focus on genes of interest. Codes are available on GitHub (https://github.com/AlinaDette/Differential-expression-of-Neuropilin-1-NRP1-and-Neuropilin-2-NRP2-in-fatal-COVID-19↗).

We performed FISH on 1-µm TMA sections of spleen and lymph node from two COVID-19 cases (patients 2 and 3, Table S1A) with the RNAscope Multiplex Fluorescent Reagent Kit v2 assay (Advanced Cell Diagnostics, Inc., Hayward, CA, USA), following the protocol previously described by reference 31.

A total of 50,000 cells per 12-mm cover slips were seeded in complete media. After at least 6 h, cells were treated according to the experimental condition and incubated overnight. Immunofluorescence was performed as previously described (73) with the following modifications: cells were fixed on ice for 10 min with 4% formaldehyde in PBS and treated with 1% BSA and 0.5% Triton X-100 in PBS for 20 min prior to incubation with primary and secondary antibodies (Table S5). NRP1 and NRP2 on HEK293 cells were detected using specific antibodies (EPR3113 and EPR23808-72, respectively; Abcam Ltd., Cambridge, UK). HEK293 cells not expressing the NRP2 construct were used as internal negative controls for binding specificity. C29F4 antibody (Cell Signaling Technology, Leiden, Netherlands) was used for HA-tagged spike protein fragments. Rabbit isotype controls were purchased from Cell Signaling Technology. Controls are depicted in Fig. S4. Cell membrane/F-actin staining was performed using phalloidin (A30104, Thermo Fisher Scientific, Langerwehe, Germany) and cells were embedded using Fluoromount-G mounting medium with DAPI (Thermo Fisher Scientific).