Type I Interferon Response and Vascular Alteration in Chilblain-like Lesions During the COVID-19 Outbreak

L. Frumholtz; J.-D. Bouaziz; M. Battistella; J. Hadjadj; R. Chocron; D. Bengoufa; H. Le Buanec; L. Barnabei; S. Meynier; O. Schwartz; L. Grzelak; N. Smith; B. Charbit; D. Duffy; N. Yatim; A. Calugareanu; A. Philippe; C.L. Guerin; B. Joly; V. Siguret; L. Jaume; H. Bachelez; M. Bagot; F. Rieux-Laucat; S. Maylin; J. Legoff; C. Delaugerre; N. Gendron; D.M. Smadja; C. Cassius


The British Journal of Dermatology. 2021;185(6):1176-1185. 

In This Article


Patient Characteristics

Fifty patients were included; their characteristics are described in Table S1 (see Supporting Information). Twenty-nine (58%) had suggestive extracutaneous COVID-19 symptoms, including asthenia (n = 14), fever (n = 11), upper airway and ear–nose–throat symptoms (n = 16), cough (n = 9), dyspnoea (n = 2) and anosmia (n = 1). Twenty (40%) had been in close contact with a person with confirmed COVID-19. The median duration between suggestive extracutaneous symptoms (when present) and CLL was 7 days [interquartile range (IQR) 2–16]. The median duration between suggestive extracutaneous symptoms and study inclusion was 24 days (IQR 16–31). The median duration between onset of CLL and study inclusion was 15 days (IQR 10–21).

Toes and fingers were involved in 86% and 24% of patients, respectively (Figure S1; see Supporting Information). Other cutaneous manifestations concurrent with CLL were extensive livedo (n = 2; 4%), arthralgia (n = 1; 2%), erythematous papules on scars (n = 2; 4%), maculopapular exanthema (n = 4; 8%) and urticaria (n = 1; 2%). CLL remission at day 15 was observed in only 16 patients (32%). Seven patients (14%) developed a new CLL flare in the meantime, and 10 patients (20%) had persistent livedo or peripheral vascular disease at day 15.

Association With SARS-CoV-2

Real-time RT-PCR testing for SARS-CoV-2 was performed using a nasopharyngeal swab at inclusion for all patients with CLL (n = 50), and in the skin (n = 6) and the stools (n = 3) of some patients. RT-PCR was negative for all samples tested.

COVID-19 serological tests were performed using three different techniques at inclusion and 14 days later: (i) IgG COVID Abbott Architect test (Abbott Laboratories, Libertyville, IL, USA); (ii) IgG and IgA enzyme-linked immunosorbent assay (ELISA) nucleocapsid COVID EUROIMMUN tests (EUROIMMUN, Lübeck, Germany); and (iii) flow spike IgA and IgG detection.[16] The serological results were compared with the data from patients with RT-PCR-confirmed mild COVID-19. The serological tests were all negative in the CLL group, except for four positive and four doubtful IgA ELISA anti-SARS-CoV-2 tests at the first visit (Table S2; see Supporting Information). All samples from the comparator group with confirmed COVID-19 were positive.

Chilblain-like Lesions are Characterized by Immune Cytotoxic Infiltration, Interferon Polarization and Endothelial Alteration in the Skin

To assess the pathophysiology of CLL we first studied all skin histological characteristics of 13 patients with CLL, compared firstly with 13 patients with SC before the COVID-19 pandemic and secondly with HC. All CLL and SC samples showed lymphocytic infiltration around superficial and deep dermal blood vessels. Other histological patterns were frequent in both CLL and SC, namely dense perivascular lymphocytic infiltrate, lymphocytic acrosyringitis, lymphocytes around the sweat gland coil, papillary oedema, extravasation of red blood cells, presence of CD123-positive plasmacytoid dendritic cells, presence of plump endothelial cells, and presence of CD61-positive platelet microthrombi. CD61 staining was used to highlight platelet microthrombi. IgA staining was observed in interstitial papillary dermis only in CLL and not in SC (Table 1 and Figure 1a).

Figure 1.

Chilblain-like lesions (CLL) and seasonal chilblains (SC) display a common transcriptomic signature and histological pattern. (a) Histological sections of CLL skin samples representing from top to bottom: haematoxylin–eosin–safran (HES) showing dense perivascular lymphocytic infiltrate (scale bar = 100 μm); HES showing plump endothelial cells (scale bar = 100 μm); endovascular CD61 immunostaining (scale bar = 100 μm) and diffuse IgA immunostaining (scale bar = 250 μm). (b–d) Skin transcriptomic profiles of CLL (n = 10), SC (n = 4) and healthy controls (HC) (n = 4). (b) Two-dimensional hierarchical clustering was performed on a set of 70 genes whose average expression significantly differed between CLL, SC and HC. All values were log2 transformed and mean centred across each gene. (c) Principal component analysis of gene expression data of skin samples showing the distance between CLL, SC and HC. (d) Venn diagram comparing differently expressed genes with P < 0·05 between CLL, SC and HC.

We compared the skin immune transcriptomic signature of 10 CLL, four SC and four HC skin samples (Figure 1b–d). The three groups clustered independently with an overlap between two CLL and SC samples, suggesting a common immune pathophysiology. Among 589 genes, we identified 296 differentially expressed genes (DEGs) between CLL and HC, most of them being upregulated. Common DEGs between CLL and SC included type I and II interferon pathways (upregulation of CXCL9, CXCL10, CXCL11, IDO1, TLR1, TLR2, TLR7, TLR8, MYD88, IRF5, IFNAR2, IFI16, IRAK1, IRAK4 and NFKB1); T helper 1 cell polarization (upregulation of CXCR3 and downregulation of GATA3 and RORC); cytotoxicity markers (upregulation of granzyme A, granzyme B, granzyme K, granulysin and perforin); natural killer cell signalling (upregulation of KLRC2, KLRC1, KLRD1, KLRC3, CD96, KLRC4, KLRF1, KLRB1 and KLRG1) and other activation and regulatory immune markers (LILRB1, LAG3, TIGIT, CTLA4, FOXP3, IL2RA, BTLA, HLA-DR genes and PDCD1LG2).

Interestingly, 57 DEGs were specific to CLL, involving: (i) complement activation (upregulation in the classical pathway of C1q, C1s and C1 inhibitor, and the alternative pathway C2 and properdin, with downregulation of membrane attack complex components C5 and C6); and (ii) angiogenesis (upregulation of PDGFB and downregulation of TNFSF12) (Figure 1).

IgA Antineutrophil Cytoplasmic Antibodies and Elevated Type I Interferon Signature Characterize Systemic Immune Response in Chilblain-like Lesions

We next assessed the systemic immune response during CLL. Firstly, ANA were positive in 10 (20%) patients (titre > 160) but none had either anti-extractable nuclear antigen or anti-DNA antibodies. Three (6%) patients had significant levels of cryofibrinogen. Three (6%) patients had low levels of C3 (range 0·73–0·75g L−1) and 10 (20%) had low levels of C4 (range 0·12–0·15g L−1). Concerning ANCA, IgG ANCA were positive in four patients, with cytoplasmic fluorescence without specificity. Notably, IgA ANCA were positive in 34 of 46 (74%) patients with cytoplasmic fluorescence: four with proteinase 3 specificity, three with bactericidal/permeability increasing protein (BPI) specificity, one with lactoferrin specificity and three with cathepsin G specificity. IgA ANCA were positive in all five SC samples that could be analysed.

We next studied circulating cytokines and the blood type I IFN signature at day 0 and day 14 in CLL compared with patients with non-severe COVID-19, patients with SAVI, and HC. As shown in Figure 2(a), the type I IFN gene signature in whole blood was significantly higher in CLL than in HC. Three patients reached the level observed in patients with SAVI. CLL samples also exhibited a peculiar cytokine profile with a significant elevation of IFN-α2 and nonsignificant elevation of IFN-γ compared with HC, and a significant decrease of IL-10 (Figure 2b). Other cytokines (IL-6, TNF-α and IL-17A) were similar in both groups (Figure 2c). When initially elevated, cytokine levels normalized at day 14.

Figure 2.

Patients with chilblain-like lesions (CLL) display a systemic immune response with an enhanced type I interferon (IFN) signature. (a) IFN-stimulated gene (ISG) score based on the expression of six genes (IFI44L, IFI27, RSAD2, SIGLEC1, IFIT1 and ISG15) measured with quantitative reverse-transcriptase polymerase chain reaction in whole blood cells from patients with CLL (n = 57), healthy controls (HC) (n = 20) and patients with STING-associated vasculopathy in infancy (SAVI) (n = 4). (b, c) Serum cytokine concentrations were measured in HC (n = 7), patients with CLL at day 0 (D0) (n = 50) and day 14 (D14) (n = 7) and patients with COVID-19 (n = 4). The data are presented as scattered dot plots; the horizontal bar represents the median and the whiskers represent the interquartile range. P-values were calculated using the Wilcoxon–Mann–Whitney test; P-values < 0·05 were considered significant. IL, interleukin; TNF, tumour necrosis factor. *P < 0·05; where not specified (ns), differences are not statistically significant.

Systemic Endothelial Activation

Lastly, we studied the systemic modifications in haemostasis and endothelial activation during CLL. The PT ratio and aPTT were normal in all patients. Only three patients (6%) had a moderate increase in D-dimer levels (510, 560 and 1680 ng mL−1). Antithrombin, and protein C and protein S activity levels were within the normal range in all patients, except for three (6%) with slightly decreased protein S activity (45, 48 and 49 IU dL−1). Five patients (10%) had transient positive lupus anticoagulant, which was negative 15 days later. One patient (2%) had isolated IgM anticardiolipin positivity and one (2%) had isolated IgM anti-β2-glycoprotein-I positivity.

We next quantified four biomarkers related to endothelial dysfunction or activation (Figure 3a) and four biomarkers related to angiogenesis or endothelial progenitor cell mobilization (Figure 3b). We found a significant increase in all biomarkers of endothelial dysfunction in CCL compared with HC. Among angiogenic-related biomarkers, VEGF-A, VEGFR-2 and c-Kit were significantly increased, while b-FGF was similar. Kinetics of endothelial dysfunction biomarkers (day 0 and day 14 after first sampling) suggested normalization of endothelial activation with time (Table S3; see Supporting Information). However, there was no association between the level of endothelial markers and recovery in patients with CLL.

Figure 3.

Patients with chilblain-like lesions (CLL) display circulating markers of endothelial injury. Soluble serum markers (a) of endothelial dysfunction or activation and (b) related to angiogenesis or endothelial progenitor cell mobilization were measured in patients with CLL at day 0 (D0) (n = 34) and day 14 (D14) (n = 34), patients with COVID-19 (n = 19) and healthy controls (HC) (n = 34). The data are presented as scattered dot plots; the horizontal bar represents the median, and the whiskers represent the interquartile range. P-values were calculated using the Wilcoxon–Mann–Whitney test; P-values < 0·05 were considered significant. bFGF, basic fibroblast growth factor; VEGF, vascular endothelial growth factor; VEGFR, VEGF receptor. *P < 0·05, **P < 0·01, ****P < 0·0001.

Increase of endothelial markers was associated with the delay between blood sampling and CLL onset (Table S4; see Supporting Information). In particular, angiopoietin-1, angiopoietin-2 and VEGF-A were increased in patients with CLL sampled > 20 days after skin lesions had started ('late' CLL) compared with patients sampled at earlier timepoints ('early' CLL) (P = 0·05, 0·007 and 0·003, respectively). Similarly, CECs were also upregulated in 'late' vs. 'early' CLL (P = 0·03).