immune dysregulation in common variable
immunodeficiency
Jonas Hultberg, MD,a,bJan Ernerudh, MD, PhD,bMarie Larsson, PhD,aA˚ sa Nilsdotter-Augustinsson, MD, PhD,c* and Sofia Nystr€om, MD, PhDa,b
* Link€oping, Sweden
GRAPHICAL ABSTRACT
Plasma protein profiling reflects Th1-driven immune
dysregulation in common variable immunodeficiency
CVID (n=29) HC (n=36)
Immune dysregulation
PEA (145 plasma proteins)
HC, healthy controls; CVID, common variable immunodeficiency; PEA, proximity extension assay; ILD, interstitial lung disease
mmune dysregulationn
Data processing CVID-B n=17
CD21low B-cells >20% (6 of 17) CVID-A n=11 Autoimmunity (7 of 11) C CD21low B-cells >20% (8 of 11) >20% (6 of > cells >ll > % (6 o 17>>20%>2 % (20%%%% 6 oof ow VID BB ILD (5 of 11) ILD (2 of 17) Autoimmunity (5 of 17) 21l B D D D2222121l 2022 C C C CD C CD C C 21lolo D 1 D l V D21 VID 21 CV D2 w BB B B B 77 ll 77 ll 77 177 1 n n B B B B CXCL10 CXCL9 IFN-CXCL11 CD40L IL12 CCL3 LAMP3 IL8 CD40 CXCL1 PDL1 CCL4
Background: Common variable immunodeficiency (CVID) is a disorder characterized by antibody deficiency. A significant fraction of the patients suffer from immune dysregulation, which leads to increased morbidity and mortality. The pathogenesis of this condition is poorly understood.
Objective: Our aim was to find out whether the plasma protein signature in CVID is associated with clinical characteristics and lymphocyte aberrations.
Methods: A highly sensitive proximity extension assay was used for targeted profiling of 145 plasma proteins in 29 patients with
CVID. Phenotyping of peripheral lymphocytes was done by flow cytometry. The findings were correlated with the burden of immune dysregulation.
Results: Unsupervised clustering of plasma protein profiles identified 2 distinct groups of patients with CVID that differed significantly in terms of the degree of complications due to immune dysregulation and in terms of the frequency of activated B- and T-cell subpopulations. Pathway analysis identified IFN-g and IL-1b as the top enriched upstream regulators associated with higher grade of immune
Fromathe Division of Molecular Virology,bthe Department of Clinical Immunology and
Transfusions Medicine, andcthe Department of Infectious Diseases, and Department of Biomedicine and Clinical Sciences, Link€oping University, Link€oping. *These authors are senior authors and contributed equally to this work. Supported by ALF-grants, Region €Osterg€otland, Sweden.
Disclosure of potential conflict of interest: The authors declare that they have no relevant conflicts of interest.
Received for publication July 25, 2019; revised January 14, 2020; accepted for publica-tion January 17, 2020.
Available online February 11, 2020.
Corresponding author: Sofia Nystr€om, MD, PhD, Clinical Immunology and Transfusion Medicine and Department of Biomedicine and Clinical Sciences, University Hospital, S-58185 Link€oping, Sweden. E-mail:sofia.c.nystrom@liu.se.
The CrossMark symbol notifies online readers when updates have been made to the article such as errata or minor corrections
0091-6749
Ó 2020 The Authors. Published by Elsevier Inc. on behalf of the American Academy of Allergy, Asthma & Immunology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
https://doi.org/10.1016/j.jaci.2020.01.046
dysregulation. In addition, CVID was found to be associated with increased plasma levels of the B-cell–attracting chemokine CXCL13.
Conclusion: Clustering based on plasma protein profiles delineated a subgroup of patients with CVID with activated T cells and clinical complications due to immune dysregulation. Thus, data indicate that CVID-associated immune dysregulation is a TH1-mediated inflammatory process driven by the IFN-g pathway. (J Allergy Clin Immunol 2020;146:417-28.)
Key words: CVID, immunodeficiency, lymphoproliferation, immune dysregulation, proximity extension immunoassay, plasma proteomics
Common variable immunodeficiency (CVID) is a heteroge-neous disorder characterized by hypogammaglobinemia and poor vaccine responses.1-3A number of genetic autosomal variants that predispose for B-cell failure have been identified in familial CVID.4However, the vast majority of CVID cases are sporadic and match a polygenic model.5-8Immunoglobulin replacement therapy reduces the number of bacterial infections in CVID, but a significant fraction of the patients with CVID develop noninfec-tious complications that are the main drivers of morbidity and mortality.9Noninfectious complications involve immune dysre-gulation, including organ-specific autoimmunity, polyclonal lymphoid infiltration of nonlymphoid tissue, and lymphoid hyper-plasia.10Patients with low numbers of class-switched memory B cells and abnormalities in T cells, including low frequencies of naive CD4 THcells and an immunosenescent T-cell phenotype,
are at higher risk of complications related to immune dysregula-tion.11-13Another potential predictor of immune dysregulation in CVID is increased frequency of B cells with low expression of CD21 (CD21-low), a distinct population of activated B cells with reduced effector functions.14,15The population closely re-sembles CD11c1 B cells, an atypical B-cell population that is maintained under conditions of chronic antigen stimulation char-acterized by aberrant type 1 and type 2 interferon responses, such as autoimmunity and persistent viral infections.16
Recent evidence suggests that interferon activation is involved in CVID-related immune dysregulation, as shown by whole blood transcriptional profiling and increased serum levels of IFN-g.17,18 It has also been reported that immune dysregulation and the expansion of CD21-low B cells are associated with a TH
1-biased CD4 T-cell compartment in CVID.19Deeper insight into the molecular mechanisms and putative role of interferons in CVID-associated immune dysregulation can pave the way for new treatment strategies to mitigate CVID-related immune dysregulation.
Proteomics is a promising technique to provide new diagnostic and prognostic information as well as insights into new disease mechanisms in cancer, inflammatory disease, and cardiovascular disease.20-23Prior work in CVID has used conventional immuno-assays for the measurement of restricted numbers of biomarkers in plasma. The proteomic approach can be useful in the identifi-cation of pathways involved in CVID with and without immune dysregulation; it can also provide information about the underly-ing molecular pathogenesis and provide markers for disease progression.
To characterize potential systemic inflammatory risk signa-tures in CVID we assessed a large number of established and exploratory immune-related biomarkers by using targeted
proteomics. Unbiased cluster analysis of 145 immune-related proteins in plasma, assessed by highly sensitive multiplex proximity extension assay (PEA) technology,24,25 identified a distinct group of patients with CVID with immune dysregulation, low frequencies of naive CD4 T cells, and increased frequencies of CD21-low B cells. Pathway analysis of plasma proteins showed that immune dysregulation was primarily associated with IFN-g and IL-1b. Plasma protein profiling can provide mechanistic information on the pathophysiology underlying im-mune dysregulation in CVID and can also provide sensitive bio-markers and guidance on tailored treatment strategies.
METHODS Patients
The study protocol was approved by the regional ethics committee in Link€oping (2017/214-32). All participants gave and signed consent before participation in this study. Before enrollment all patients were reevaluated for fulfilling the diagnostic criteria of CVID. In all, 29 patients with CVID (19 males and 10 females aged 18-77 years [median age 53 years]) were enrolled in the study. Clinical data were collected from their medical records and covered the time of the study start to 10 years back. An evaluation document, based on division into distinct clinical phenotypes, was used to summarize the medical records of each patient. In brief, patients were assessed for the presence of 3 noninfectious complications: autoimmunity, gastroenteropathy, and lymphoproliferation (seeTable E1in this article’s Online Repository at
www.jacionline.org). Each item of complications was scored 0 to 4; hence, the sum formed a disease severity score between 0 and 12. Patients were also given an infection score between 0 and 4 based on the number of sus-pected infections per year after initiation of IgG substitution therapy, as docu-mented in the medical records of the patients. Information about B-cell subsets was retrieved from the clinical records of the patients. CVID was classified ac-cording to the Freiburg criteria based on the proportions of B-cell subsets of peripheral blood lymphocytes.26Blood donors (27 males and 9 females) aged 18 to 67 years (median age 50 years) served as healthy controls. Blood samples were obtained in EDTA tubes from patients and healthy controls. One blood sample was used for plasma isolation and another was used for im-mediate lymphocyte immunophenotyping. All samples were collected in a standardized manner by experienced staff using the same technique (Vacu-tainer [BD Biosciences, Franklin Lakes, NJ]) and tubes. Samples were kept at room temperature up to 5 hours before plasma was separated.
Plasma protein profiling by proximity extension assay
Plasma samples were collected, centrifuged, and stored at2808C until further processing. Aliquots were analyzed with the Proseek multiplex assay (Olink Proteomics, Uppsala, Sweden), a PEA technology.24,25Two panels
Abbreviations used
AUC: Area under the curve
CRTAM: Cytotoxic and regulatory T-cell molecule CVID: Common variable immunodeficiency CXCL: Chemokine C-X-C motif ligand FcRL6: Fc receptor-like protein 6
HLA-DR: Human leukocyte antigen–antigen D–related IPA: Ingenuity Pathway Analysis
NF-kB: Nuclear factor-kB Pcorr: Corrected P value
PCA: Principal component analysis PEA: Proximity extension assay PDCD1: Primary death protein 1
(Immune Response and Immuno-oncology) of 92 proteins each were used to detect 173 unique immune-related proteins (seeTables E2andE3in this arti-cle’s Online Repository atwww.jacionline.org). IFN-g level was analyzed with the PEA Inflammation panel (Olink Preoteomics, 95302). Frozen ali-quots of samples were sent to the Clinical Biomarkers Facility, Science for Life Laboratory, Uppsala University (Uppsala, Sweden) and analyzed accord-ing to its standard operataccord-ing procedures. Data from the 2 data sets were merged. Markers with a call rate less than 75% were discarded and not included in subsequent analysis. Any missing values because of low concen-tration were set to a value between the LOD and 0 (seeTable E2for details of the LOD and call rates).
Immune phenotyping by flow cytometry
Absolute T-cell numbers were determined, and phenotyping of T-cell subsets in peripheral blood samples was performed by flow cytometry with the use of Trucount tubes (BD Biosciences). Briefly, antibodies were added to whole blood and samples were lysed with FACS lysing solution (BD Biosciences) before data were acquired by using FACS Canto II (BD Biosciences). Kaluza flow cytometry software version 1.5 (Beckman Coulter, Brea, Calif) was used for data analysis. Lists of the mAbs used and the gating strategy of T-cell subsets are available inTable E4andFig E1in this article’s Online Repository atwww.jacionline.org.
Statistical analysis
Statistical analysis of plasma protein profiling was carried out by using a 2-tailed t test in Microsoft Excel. The P value for each protein was adjusted for multiple comparisons by using the corrected P value (Pcorr) obtained by the Benjamini-Hochberg procedure. Ingenuity Pathway Analysis (IPA) (Qia-gen, Hilden Germany) of canonic pathway enrichment and upstream regula-tors was performed by using the differences in mean plasma protein levels that were considered statistically different between groups; that is, mean protein levels that differed by at least a log2 value of 0.3 but no more than a negative log2 value of –0.3 between groups in combination with a value of Pcorrless than .01 (patients with CVID vs healthy controls) or Pcorr less than .05 (CVID group A vs CVID group B). Prediction of affected path-ways and activated upstream regulators was based on the overlap of the differently expressed proteins and proteins associated with the pathway or the downstream targets of the regulators, as determined by using the Fisher exact test. Differences in the levels of the downstream targets generate z scores ranging from14 to –4, with a positive z score predicting activation and a negative z score predicting inactivation of the upstream regulator. A z score of12 or higher was considered significantly activated. Phyton was used for principal component analysis (PCA). Cluster analysis and heatmaps were constructed with the use of Heatmapper27and Venn diagrams were constructed by using the tool freely provided by The Bioinformatics and Evolutionary Genomics research unit of Ghent University (Ghent, Belgium) (http://bioinformatics.psb.ugent.be/webtools/Venn/). In addition, results were analyzed by using GraphPad Prism 7 (GraphPad Software, San Diego, Calif) with 1-way ANOVA followed by Tukey postest and Pearson correla-tion analysis and the Mann-Whitney U test or Spearman correlacorrela-tion test for nonparametric data. Microsoft Excel was used for chi-square analysis of categoric data. Receiver operating characteristic curves and area under the curve (AUC) were calculated for plasma proteins increased in CVID group A compared with in CVID group B by using Metaboanalyst 4.0.28The clas-sification performance (sensitivity and specificity) of the proteins was also evaluated. A P value less than .05 was considered significant unless other-wise stated.
RESULTS
Plasma protein profiles of patients with CVID To assess the plasma protein signature associated with CVID, we compared the protein levels in 29 patients with CVID and 37 healthy controls by using plasma obtained by PEA. Of 173
analyzed proteins, 145 were detectable in more than 75% of the samples and were included in further analysis. Among the proteins, the mean plasma levels of 72 proteins differed significantly, with a value of Pcorrless than .01 between patients
with CVID and healthy controls (seeTable E5 in this article’s Online Repository atwww.jacionline.org). Most of the signifi-cantly affected proteins in patients with CVID were upregu-lated, and only a minority were downregulated. Normalized protein expression of the 6 extracellular proteins and the 4 membrane proteins with the most significant difference in mean expression between patients with CVID and controls are shown (Fig 1, A and B). The mean plasma levels of these pro-teins were 2.8 to 4 times higher (log25 1.5-2.0) in the CVID
group than in the controls. Chemokine C-X-C motif ligand 13 (CXCL13), also known as B-cell lymphocyte chemoattractant factor, was the soluble factor with the highest levels in patients with CVID compared with in the controls (Fig 1, A). In addi-tion, CXCL9 and CXCL10 were upregulated in CVID compared with in the controls (Fig 1, A). The top upregulated membrane-bound proteins in patients with CVID compared with in the controls included 2 members of the tumor necrosis factor receptor superfamily (TNFRSF), the immune checkpoint primary death protein 1 (PDCD1), and the cytotoxic and regu-latory T-cell molecule (CRTAM) (Fig1, B). Enrichment analysis by IPA revealed significant overlap of the differently expressed plasma proteins in patients with CVID with proteins associated with several canonic pathways. The 3 pathways with the highest probability to be affected in patients with CVID when compared with the controls were communication between innate and adaptive immune cells (P < 10–13), the TH1 pathway (P < 10– 12), and the T
H1 and TH2 activation pathways (P < 10–12)
(see Fig E2 in this article’s Online Repository at www. jacionline.org). Further IPA analysis predicted IL-36B as the top upstream regulator of CVID plasma protein profiles (P < 10–16) based on increased levels of the following proteins: IL-6, CXCL10, chemokine C-C motif ligand CCL20, IL-18, CD83, CXCL8, CXCL1, and CD40, proteins that are all regu-lated by IL-36B (seeTable E5in this article’s Online Reposi-tory atwww.jacionline.org).
The plasma protein profiles identified a distinct group of patients with CVID
Unsupervised hierarchic clustering analysis identified 1 group of patients with CVID (11 of 29) that clustered separately from the other study subjects (Fig 2, A) (ie, CVID group A). Of the remaining patients, 17 who intermingled with the controls to some extent were classified as CVID group B. One patient was not allocated to any of the groups (Fig 2, A). Characteriza-tion of CVID group A and CVID group B showed that the fre-quency of atypical B cells, with low expression of CD21, was significantly higher in CVID group A (Table I). The presence of noninfectious complications—lymphadenopathy, enteropa-thy, splenomegaly, or lymphocytic infiltration with or without autoimmunity—was significantly higher in group A than in group B, but there was no difference in the frequency of docu-mented infections between groups (Fig 2, B). Only 4 of 28 pa-tients had infection only, and they all clustered in group B. The number of patients classified as CVID Freiburg class 1a, which indicates an increased risk of noninfectious complica-tions, was significantly higher in CVID group A (Fig 2, C).
Histopathologic signs of interstitial lung disease were also more common in CVID group A (Table I). Accordingly, plasma pro-tein pattern profiling can identify patients with CVID with a se-vere clinical phenotype characterized by immune dysregulation. Female-to-male ratio, age, and a positive family history of pri-mary immunodeficiency were equally distributed between groups (Table I). There was no significant difference in the pres-ence of gastroenteropathy between groups. However, IPA indi-cated enrichment of plasma proteins linked to hepatic fibrosis and hepatic stellate activation pathways (P value 10–11) in
CVID group A versus in CVID group B. There was overlap be-tween the top enriched pathways in CVID group A and those in CVID group B and between the pathways in patients with CVID and those in healthy controls (seeFig E2in this article’s Online Repository atwww.jacionline.org). Four patients in CVID group B had a high disease score. A PCA loadings plot showed that decreased levels of CXCL5 had strong influence on the separa-tion of these patients from CVID group A patients in addisepara-tion to decreased levels of CXCR3 ligands (seeFig E3in this article’s Online Repository atwww.jacionline.org).
FIG 1. Plasma protein profiles of patients with CVID. Plasma levels of the top 6 deregulated extracellular proteins (A) and top 4 deregulated membrane proteins (B) in patients with CVID compared with in healthy controls (HCs). Data are presented as normalized protein expression (NPX) in log2scale. Bars and error bars
indicate mean and 95% CI. Pcorrrepresents theP value (determined by t test) corrected for multiple
The CVID plasma protein profiles are associated with increased T-cell activation
Perturbation of the CD4 TH subset is associated with an
increased risk of noninfectious complications in CVID.29 We therefore assessed the extended phenotype of peripheral T cells in patients with CVID and controls. The number of CD4 T cells was decreased in CVID group B, which was characterized by the milder phenotype, compared with in healthy controls. In both CVID groups there was a decrease in the frequencies of naive CD4 T cells compared with in the controls, and the frequency was significantly lower in CVID group A than in CVID group B (Fig 3, A). This is in line with previous reports that low numbers of naive CD4 T cells are associated with noninfectious complications in CVID.12The frequency of effector memory CD4 T cells was increased in CVID group A compared with in CVID group B and in healthy controls (Fig 3, A). In addition, the frequency of CD4 T cells expressing the activation marker human leukocyte antigen–antigen D–related (HLA-DR) was higher in CVID group A than in CVID group B (Fig 3, B). Increased levels of CD4 T
cells expressing the immune checkpoint PDCD1 were found in both CVID groups, but the levels of PDCD11CD4 T cells were higher in CVID group A than in CVID group B (Fig 3, B). Plasma levels of PDCD1 correlated (r5 0.57; P 5 .001) with the fre-quency of PDCD11CD41T cells (seeFig E4in this article’s On-line Repository at www.jacionline.org). The levels of central memory (CD45RA–and CD62L1) and terminally differentiated RA1 (CD45RA1and CD62L–; terminally differentiated RA1) CD4 T cells did not differ between groups.
The number of CD8 T cells was higher in CVID group A than in CVID group B and in the healthy controls (Fig 3, C). There was also a decrease in the frequencies of naive CD8 T cells in CVID group A and CVID group B compared with in the controls. CD8 T cells in CVID group A (ie, patients characterized by a more severe clinical phenotype) had a higher expression of the activation marker HLA-DR than did those in CVID group B and healthy controls (Fig 3, D). Increased levels of CD8 T cells with expres-sion of PDCD1 was restricted to CVID group B when compared with the controls (Fig 3, D). There was no difference in the levels
B
Disease score
Infection
score
Immune-dysregulation
y
g
Infectious complications
Infectious complications
C
Freiburg class CVID-A CVID-B P-value#
1a N=8 N=4 0.010
1b N=1 N=5 ns
2 N=2 N=8 ns
A
FIG 2. The plasma protein signature identified a distinct group of patients with CVID. A, Tree diagram of un-supervised cluster analysis of the plasma protein profiles of patients in CVID group A (CVID-A) (A), CVID group B (CVID-B) (B), and healthy controls (C). Tree diagram indicates similarity in plasma protein signature by common branches. The star represents 1 patient with CVID not allocated to either group CVID-A or group CVID-B. B, Disease and infection score in group CVID-A and group CVID-B. A and B, Patients in group CVID-B with a disease score of 7 are indicated in red. C, The table presents the classification of patients in groups CVID-A and CVID-B, according to the Freiburg classification scheme. **P < .01 (Mann-Whitney U test)#
Chi-square test.ns, Not significant.
of the CD8 memory T-cell subsets (central memory, effector memory, and terminally differentiated RA1) between groups.
Noninfectious complications are associated with a plasma protein profile under the influence of IFN-g Plasma proteins were considered upregulated, with a fold change of 1.3 or more (linearized value) and significantly different if Pcorr was less than .05 when CVID group A was
compared with CVID group B (seeTable E6in this article’s On-line Repository atwww.jacionline.org). In total, 84 proteins were upregulated and no proteins were downregulated in CVID group A, when compared with CVID group B. The plasma profiles of the patients in CVID group A were dominated by increased levels of 3 CXCR3 IFN-g–inducible ligands: CXCL9, CXCL10 and CXCL11 (Table II). The top upregulated CVID group A– associated membrane proteins include Fc receptor-like protein (FcRL) 6, TNFRSF4, and TNFRSF9 (Table II). As an unbiased approach, we next used the IPA to analyze the upstream regulators of the protein signature that characterized patients with CVID with a higher incidence of noninfectious complications, namely, group A. IFN-g, IL-1b, and nuclear factor kappa-kB (NF-kB) were identified as the top 3 enriched upstream regulators in CVID group A when compared with CVID group B (Fig 4, A). In total, 38 plasma proteins regulated by IFN-g, IL-1, or NF-kB, were present at higher levels in patients in CVID group A. IFN-g was the predicted regulator of 24 proteins, and IL-1b and NF-kB regulated 20 proteins each (Fig 4, A). Of the 38 pro-teins, 10 were unique for IFN-g, 7 were unique for IL-1b, and 4 were unique for NF-kB, and 9 proteins were shared by all up-stream regulators (Fig 4, B). CXCL9, CXCL10, and CXCL11 were shared between all 3 regulators. Next, a PCA was performed for factors regulated by IFN-g, including for all patients in CVID group A and CVID group B and the healthy controls. Dimensional reduction of protein levels under the regulation of IFN-g showed a clear separation of CVID group A and CVID group B when the
data were projected to PC1 and PC2 (Fig 4, D). The relative plasma levels of the markers are shown in heatmaps (Fig 4, D). IFN-g was initially detected in less than 15% of the analyzed sam-ples and was therefore not included in the pathway analysis. How-ever, a PEA assay with improved sensitivity for IFN-g showed significantly increased plasma levels in patients in CVID group A compared with in patients in CVID group B and healthy con-trols (Fig 4, C).
Plasma CXCL10 is a potential biomarker of immune dysregulation in CVID
The correlation of CXCL9, CXCL10, and CXCL11 plasma levels with naive CD4 T cells, PDCD11CD4 T cells, and HLA-DR1CD8 T cells was tested in patients with CVID (Fig 5, A). Plasma levels of CXCL9 correlated negatively with naive CD4 T cells and positively with proportions of PDCD11of CD4 T cells and proportions of HLA-DR1of CD8 T cells (Fig 5, A). CXCL10 and CXCL11 also showed significant positive correlations (r > 0.50 and P < .01) with the proportions of PDCD11of CD4 T cells and HLA-DR1of CD8 T cells (Fig 5, A). In addition, CXCL10 correlated with the frequencies of CD21-low B cells (Fig 5, B). To investigate the performance of CXCL9, CXCL10, and CXCL11 in classifying patients as belonging to CVID group A or CVID group B, we constructed receiver operating character-istic curves. Plasma levels of CXCL10 performed well in the clas-sification of CVID group A and CVID group B, with an AUC of 0.93 (95% CI5 0.762-1.0) (Fig 5, C). Sensitivity was 90% and specificity was 90% at a cutoff of more than 9.7 normalized pro-tein expression units (Fig 5, D), which equals an analytic level of approximately 1 ng/mL (IMMUNO-ONCOLOGY, v.3101;www. olink.com/downloads). CXCL9 and CXCL11, both with an AUC greater than 0.90, performed less well and resulted in misclassifi-cation of 2 patients in CVID group A (data not shown). CXCL10 showed a significant correlation with the immune dysregulation disease score (r5 0.52; P 5 .004; Spearman correlation), and
TABLE I. Demographics and clinical characteristics of patients with CVID
Characteristic All patients with CVID Group A Group B P value
No. of patients 29* 11* 17* —
Sex (female/male) 10/19 5/11 5/17 .39
Age at study start, y (range) 53 (21-77) 53 (22-74) 55 (21-77) .55
Years of follow-up 7.25 3.5 10 <.01
Plasma IgG level, g/L (range) 8.7 (0.3-12) 7.5 (0.3-11) 9.0 (1.4-12) .13
Lymphadenopathy 14/29 9/11 4/17 <.01
Splenomegaly 15/29 8/11 7/17 .10
Enteropathy 18/29 9/11 8/17 .07
Autoimmunity 13/29 7/11 5/17 .07
Interstitial lung diseaseà 8/29 5/11 2/17 <.05
Familial cases 7/29 2/11 4/17 .74
Lymphocyte level (3 109/L)§ 1.4 (0.4-4.4) 2.1 (1.1-4.4) 1.2 (0.4-2.3) <.05 T-cell level (3 109
/L)§ 1.1 (0.2-3.4) 1.5 (0.8-3.4) 0.9 (0.2-2.0) <.05
B-cell level (cells/mL)§ 102 (15-615) 118 (15-615) 99 (23-553) .64
B-cell subsets (% of CD19)§
Switched memory 3.1% (0.1%-24%) 1.1% (0.2%-8.8%) 3.8% (0.1%-24%) .35
CD21-low 23% (0.7%-57%) 30% (10%-57%) 16% (0.7%-45%) <.05
Medians and ranges are shown for cell frequencies. Statistical analysis compared group A with group B. The Mann-Whitney U test was used for statistical analysis unless otherwise indicated. A P value less than .05 was considered significant.
*One patient was not allocated to any of the groups. Chi-square test.
àIncluded the following histopathologic entities: granulomatous lymphocytic interstitial lung disease, pulmonary sarcoidosis, cryptogenic organizing pneumonia, and pulmonary immune proliferation.
§Reference values: lymphocytes (3 109/L), 1.1 to 4.8; T cells (
FIG 3. The plasma protein signature reflects increased T-cell activation. Phenotype of peripheral blood lymphocytes in patients in CVID group A (CVID-A), patients in CVID group B (CVID-B), and healthy controls (HCs). Panels show CD31CD41T cells (A), CD31CD81T cells (C), and the proportions of naive (CD45RA1and CD62L1), central memory (CD45RA–
and CD62L1), effector memory (CD45RA–
and CD62L–
), and terminally differentiated RA1(TEMRA) (CD45RA1and CD62L–
) subsets. Proportions of HLA-DR1and PDCD11CD4 T cells (B) and CD81T cells (D) in the 3 groups. Error bars represent the 95% CI. *P <.05; **P <.01; ***P < .001; 1-way ANOVA and Tukey posttest).HC, Healthy control; ns, not significant.
there was no correlation of CXCL10 with the infection score (r5 0.13) (data not shown).
DISCUSSION
Although CVID is characterized by low IgG levels and recurrent infections,1,30as many as 40% to 70% of the patients have noninfectious inflammatory and autoimmune complications that are major drivers of morbidity and mortality.9,10Although previous studies have described a limited set of potential bio-markers in the blood of patients with CVID,17,31 this study is the first to investigate a broad array of immune-related proteins in plasma of patients with CVID compared with in plasma healthy controls. The plasma protein profile in patients with CVID was profoundly different from that in healthy controls, with 72 dysre-gulated immune-related proteins of the 145 detected. Increased plasma levels of IFN-g and a distinctive plasma protein signature that included elevated levels of 24 proteins regulated by IFN-g was strongly linked to CVID-associated immune dysregulation.
Unbiased clustering of plasma protein profiles separated the patients into 2 distinct groups, of which 1 group was characterized by noninfectious complications related to immune dysregulation as well as aberrant B and T-cell phenotypes that are associated with increased risk of CVID-related immune dysregula-tion.3,10,26,29Thus, the plasma protein profiling succeeded in dis-tinguishing a group of patients with CVID with more severe immune dysregulation from the other patients. There is a need to identify novel noninvasive biomarkers that can be used to pre-dict the risk for CVID-related immune dysregulation in patients. We found that plasma CXCL10 correlated with the frequencies of CD21-low B cells, PDCD1 expression of CD4 T cells, and acti-vated CD8 T cells; hence, plasma CXCL10 is a potential biomarker for noninfectious complications in CVID. Importantly, there was no correlation between CXCL10 and the infection
score; hence, it is unlikely that chronic infections drive the CXCL10 aberration. However, further investigations are required in larger cohorts to establish a set of biomarkers that can direct specific immunomodulatory therapy to align with the principles of precision medicine.
The subgroup of patients with CVID with noninfectious complications showed prominent findings of T-cell abnormalities, with reduced numbers of naive T cells, increased T-cell activa-tion, and increased numbers of CD4 effector memory cells. These findings indicate that impaired T-cell function is a factor that strongly contributes to clustering and that aberrant T-cell function is a driving force in immune dysregulation. This is also supported by the increased expression of the immune checkpoint PDCD1 in CD4 T cells. Increased CD4 expression of immune checkpoint PDCD1 has previously been reported in CVID-associated autoimmunity and as a result of increased bacterial translocation in CVID, and it is consistent with chronic T-cell activation.13,32 Plasma levels of PDCD1 correlated with the proportions of PDCD11of CD4 T cells, a finding in support of the idea that plasma protein levels reflect the phenotype of circulating cells. The CVID group with immune dysregulation also showed increased plasma levels of FcRL6, which also provides an immu-nosuppressive signal that directly represses natural killer cell cytotoxicity and effector T-cell activity.33 Notably, FcRL6 expression is increased on cytolytic T cells in HIV infection, and increased expression of FCRL6 was recently associated with acute Epstein-Barr virus infection.34,35 Thus, increased levels of both PDCD1 and FcRL6 are in line with T-cell dysregu-lation. On the other hand, levels of molecules that augment cellular immune responses, such as TNFRSF9 (4-1BB), which is expressed by activated natural killer cells and CD8 T cells,36 and TNFRSF4 (OX40), which is mainly expressed by activated CD4 T cells,37were increased in plasma of the group with im-mune dysregulation CVID. These findings are consistent with
TABLE II. Top 21 upregulated plasma proteins in patients in CVID group A compared with in patients in CVID group B
Gene symbol Uniprot-id
Difference between
CVID group A and CVID group B Pcorr Location of protein
Upstream Regulator
CXCL9 Q07325 2.09 1,2E-04 Extracellular IFN-g, IL-1b, NF-kB
CXCL11 O14625 2.09 4,2E-05 Extracellular IFN-g, IL-1b, NF-kB
CXCL10 P02778 1.86 1,4E-04 Extracellular IFN-g, IL-1b, NF-kB
CXCL5 P42830 1.63 4,6E-02 Extracellular IL-1b, NF-kB
CCL19 Q99731 1.61 6,5E-04 Extracellular
EGF P01133 1.55 4,7E-03 Extracellular
GZMH P20718 1.54 3,1E-03 Cytoplasm
CD40LG P29965 1.50 4,4E-03 Extracellular IFN-g, IL-1b
DAPP1 Q9UN19 1.49 1,1E-02 Cytoplasm
FCRL6 Q6DN72 1.45 6,3E-04 Membrane
PDCD1 Q15116 1.43 4,7E-03 Membrane
TNFRSF4 P43489 1.43 6,6E-04 Membrane NF-kB
IL12A P29459 1.41 1,7E-03 Extracellular IFN-g
EIF4G1 Q04637 1.41 5,6E-04 Cytoplasm
CXCL13 O43927 1.40 2,1E-02 Extracellular
KLRD1 Q13241 1.39 5,4E-05 Membrane
LILRB4 Q8NHJ6 1.39 4,3E-05 Membrane
CCL3 P10147 1.38 2,0E-05 Extracellular IFN-g, IL-1b, NF-kB
TNFRSF9 Q07011 1.35 6,3E-04 Membrane
SH2B3 Q9UQQ2 1.33 1,1E-02 Membrane
IL18 Q14116 1.28 9,7E-04 Extracellular IL-1b
Pcorr, P value corrected for multiple comparisons.
Difference between CVID group A and CVID group B is the difference in mean plasma protein levels between CVID group A and CVID group B expressed as normalized protein expression in log2.Upstream regulator indicates whether the protein is regulated by IFN-g, IL-1b, and/or NF-kB.
C A D
PC 1 (44%)
)
%5,
21
(
2
C
P
IFN- upstream regulator
B
Regulator
Activated in
CVID-A
Number of upregulated
targets in CVID-A
CVID-A vs -B
z-score
p-value
IFN-
␥
Yes
24
4.0
3*10
-20IL1-

Yes
20
4.0
5*10
-20NF-
B
Yes
20
4.0
2*10
-18FIG 4. Noninfectious complications are associated with a plasma protein profile under the influence of IFN-g. A, Top 3 upstream regulators of the plasma protein profiles in CVID group A (CVID-A) when compared with CVID group B (CVID-B). Pathway analysis predicted a high likelihood of activation, indicated by high z scores and low P values, based on the number of upregulated targets in group CVID-A. B, The overlap of the targets between the 3 regulators. C, Plasma levels of IFN-g. D, PCA and heatmaps of normalized pro-tein expression (NPX) of plasma of propro-teins regulated by IFN-g in group CVID-A and group CVID-B and the mean NPX of healthy controls (HCs).
FIG 5. Plasma levels of CXCR3 ligands correlated with T-cell phenotypes. A, Scatter plots of plasma protein correlation with different T-cell phenotypes. B, Receiver operating curve analysis of plasma CXCL10 for clas-sification of patients with CVID into CVID group A (CVID-A) and CVID group B (CVID-B). Sensitivity and spec-ificity at the optimal cutoff is shown. C, Box plot of CXCL10 plasma levels showing the 95% Cl and medians. The mean is indicated by a diamond, and the optimal cutoff is indicated by a red line.
an activated phenotype, as supported by increased HLA-DR expression and increased numbers of effector memory T cells. Knowledge regarding the relation between soluble membrane molecules and cellular expression or cellular turnover is sparse. Soluble ligands and receptors may also interfere with immune cell cross-talk and cell signaling pathways. Taken together, our data indicate that the unique plasma protein profile reflects the presence of T-cell activation and a complex T-cell dysfunctional-ity that are more pronounced in patients with CVID with immune dysregulation.
When integrating plasma protein profiles for pathway analysis the CVID group with immune dysregulation displayed a profile consistent with TH1 deviation because the top 3 upregulated
pro-teins in this group were CXCL9, CXCL10, and CXCL11, all of which are ligands of the CXCR3 receptor and regulated by IFN-g. Increased levels of CXCL9, CXCL10, and CXCL11, in combination with increased plasma IFN-g levels in patients in CVID group A, support the notion that CVID-related immune dysregulation is under strong influence of IFN-g, a cytokine that functions as a stimulator and modulator of TH1 immune
re-sponses. This is consistent with previous reports in which CVID-related immune dysregulation has been associated with increased IFN-g production of lymph node CD4 T cells and pe-ripheral blood CD4 T cells on stimulation.19 Also, increased serum IFN-g levels were reported in CVID with noninfectious complications as a result of the expansion of inflammatory type 3 innate lymphoid cells.17CD21-low B cells resemble autoreac-tive CD11c1B cells associated with systemic lupus erythemato-sus, and chronic viral infections have been reported to express the IFN-g inducible transcription factor Tbet and CXCR3.14,19,38 IFN-g was recently reported to augment the development of au-toreactive CD11c1B cells,39a pathophysiologic link that has also been proposed in CVID.19 We show correlations of CXCL10 with aberrant T-cell populations in CVID as well as with CD21-low B cells and suggest that the CXCR3 axis is an attractive target for specific treatment to mitigate CVID-related immune dysregulation. MDX-1100, a fully human anti-CXCL10 mAb that was well tolerated and demonstrated clinical efficacy in a phase II clinical trial of rheumatoid arthritis,40is a drug that may prove effective in the treatment of CVID-related immune dysregulation by counteracting TH1 driven T-cell
activa-tion and by preventing the expansion of CD21-low B cells. Further analysis identified activation of IFN-g, IL-1b, TNF, NF-kB, and RelA as top upstream regulators of plasma protein profiles. Collectively, these regulators control pathways that converge in canonic NF-kB activation. Dysregulation of these pathways is predicted to mainly affect cell-cell interactions and cellular movement. Significant functional annotations include activation of leukocytes, leukocyte migration, inflammatory response, and activation of lymphocytes. Interestingly NFKB2 and NFKB1 haploinsufficiency have been associated with CVID-like disease.41,42
In addition to distinguishing the subgroups of patients with CVID, the plasma protein profiling also revealed aberrant expression that differed between the whole CVID group and the controls. A novel finding was the increased level of the B-cell– attracting chemokine CXCL13 in patients with CVID. CXCL13 is the ligand of CXCR5 and controls the formation of germinal centers.43CXCL13 is mainly produced by follicular dendritic cells and follicular THcells in the lymphoid tissue44,45and has
been suggested to be a plasma biomarker of germinal center
activity.46Increased plasma levels of CXCL13 may reflect aber-rant germinal center formation that has been reported in CVID.47Increased levels of IL-12A, CXCL9, and PDCD1 in pa-tients with CVID are consistent with previous findings.31,32,48 Pathway analysis predicted IL-36B to be the top upstream regu-lator of CVID plasma profiles. IL-36B is a poorly understood immune-activating cytokine belonging to the IL-1 family that has been suggested to play an important role in host defense by counteracting microbial immune evasion.49
Our study has some limitations. The first is the restricted number of patients included, especially with regard to the heterogeneity of the clinical manifestations of CVID. On the other hand, the number was sufficient for an unbiased approach of proteomic analysis to reveal the distinct findings that were obtained by the protein profiling. The restricted number of patients was a weakness of the evaluation of potential plasma biomarkers because it was not feasible to split the observations into training and test sets. However, we show that a single marker can perform well in predicting immune dysregulation. Second, a targeted approach was used for the proteomic studies, which does not give a complete picture of the plasma protein profiles. On the other hand, the PEA combines high sensitivity (by PCR amplification) and high specificity (by dual antibodies against different epitopes), which are needed for safe detection of low-abundant proteins, which is more difficult with mass spectrometry techniques. Finally, this is an explorative study, and an alternative study design is required to dissect the mechanistic role of each pathway in the development of CVID-related immune dysregulation.
To our knowledge, this study represents the first study of CVID based on plasma proteomics. The unbiased use of the plasma protein profiles demonstrates novel proteins and associated pathways that likely contribute to the development of CVID-related immune dysregulation. The plasma protein profiles associated with immune dysregulation are under strong influence of IFN-g signaling, with increased levels of TH1 chemokines.
Increased numbers of CD8 T cells and increased T-cell activation can also be explained by IFN-g–driven inflammation. Given the heterogeneity of the clinical manifestations of CVID, plasma proteomics provide a tool for distinguishing patients with CVID, who are at risk of increased morbidity, and it could also provide information of perturbed pathways available for specific immunotherapy.
We would like to acknowledge the support of the Clinical Biomarker Facility at SciLifeLab Sweden for providing assistance in protein analyses. The study was financially supported by ALF Grants, Region €Osterg€otland.
Clinical implications: Profiling of plasma protein identifies bio-markers associated with immune dysregulation in CVID that may help in delineating the underlying molecular mechanisms and pave the way for tailored treatment strategies.
REFERENCES
1.Ameratunga R, Brewerton M, Slade C, Jordan A, Gillis D, Steele R, et al. Com-parison of diagnostic criteria for common variable immunodeficiency disorder. Front Immunol 2014;5:415.
2.Cunningham-Rundles C, Bodian C. Common variable immunodeficiency: clinical and immunological features of 248 patients. Clin Immunol 1999;92:34-48. 3.Wehr C, Kivioja T, Schmitt C, Ferry B, Witte T, Eren E, et al. The EUROclass trial:
4.Picard C, Bobby Gaspar H, Al-Herz W, Bousfiha A, Casanova JL, Chatila T, et al. In-ternational Union of Immunological Societies: 2017 Primary Immunodeficiency Dis-eases Committee report on inborn errors of immunity. J Clin Immunol 2018;38:96-128. 5.Orange JS, Glessner JT, Resnick E, Sullivan KE, Lucas M, Ferry B, et al. Genome-wide association identifies diverse causes of common variable immunodeficiency. J Allergy Clin Immunol 2011;127:1360-7.e6.
6.Picard C, Al-Herz W, Bousfiha A, Casanova JL, Chatila T, Conley ME, et al. Pri-mary immunodeficiency diseases: an update on the classification from the Interna-tional Union of Immunological Societies Expert Committee for Primary Immunodeficiency 2015. J Clin Immunol 2015;35:696-726.
7.Tuijnenburg P, Lango Allen H, Burns SO, Greene D, Jansen MH, Staples E, et al. Loss-of-function nuclear factor kappaB subunit 1 (NFKB1) variants are the most common monogenic cause of common variable immunodeficiency in Europeans. J Allergy Clin Immunol 2018;142:1285-96.
8.van Schouwenburg PA, Davenport EE, Kienzler AK, Marwah I, Wright B, Lucas M, et al. Application of whole genome and RNA sequencing to investigate the genomic landscape of common variable immunodeficiency disorders. Clin Immu-nol 2015;160:301-14.
9.Resnick ES, Moshier EL, Godbold JH, Cunningham-Rundles C. Morbidity and mortality in common variable immune deficiency over 4 decades. Blood 2012; 119:1650-7.
10.Chapel H, Lucas M, Lee M, Bjorkander J, Webster D, Grimbacher B, et al. Com-mon variable immunodeficiency disorders: division into distinct clinical pheno-types. Blood 2008;112:277-86.
11.Malphettes M, Gerard L, Carmagnat M, Mouillot G, Vince N, Boutboul D, et al. Late-onset combined immune deficiency: a subset of common variable immunode-ficiency with severe T cell defect. Clin Infect Dis 2009;49:1329-38.
12.Mouillot G, Carmagnat M, Gerard L, Garnier JL, Fieschi C, Vince N, et al. B-cell and T-cell phenotypes in CVID patients correlate with the clinical phenotype of the disease. J Clin Immunol 2010;30:746-55.
13.Stuchly J, Kanderova V, Vlkova M, Hermanova I, Slamova L, Pelak O, et al. Com-mon variable immunodeficiency patients with a phenotypic profile of immunose-nescence present with thrombocytopenia. Sci Rep 2017;7:39710.
14.Isnardi I, Ng YS, Menard L, Meyers G, Saadoun D, Srdanovic I, et al. Complement receptor 2/CD21- human naive B cells contain mostly autoreactive unresponsive clones. Blood 2010;115:5026-36.
15.Rakhmanov M, Keller B, Gutenberger S, Foerster C, Hoenig M, Driessen G, et al. Circulating CD21low B cells in common variable immunodeficiency resemble tis-sue homing, innate-like B cells. Proc Natl Acad Sci U S A 2009;106:13451-6. 16.Karnell JL, Kumar V, Wang J, Wang S, Voynova E, Ettinger R. Role of CD11c(1)
T-bet(1) B cells in human health and disease. Cell Immunol 2017;321:40-5. 17.Cols M, Rahman A, Maglione PJ, Garcia-Carmona Y, Simchoni N, Ko HM, et al.
Expansion of inflammatory innate lymphoid cells in patients with common variable immune deficiency. J Allergy Clin Immunol 2016;137:1206-15.e6.
18.Park J, Munagala I, Xu H, Blankenship D, Maffucci P, Chaussabel D, et al. Inter-feron signature in the blood in inflammatory common variable immune deficiency. PLoS One 2013;8:e74893.
19.Unger S, Seidl M, van Schouwenburg P, Rakhmanov M, Bulashevska A, Frede N, et al. The TH1 phenotype of follicular helper T cells indicates an IFN-gamma-associated immune dysregulation in patients with CD21low common variable im-munodeficiency. J Allergy Clin Immunol 2018;141:730-40.
20.Carvalho VP, Grassi ML, Palma CS, Carrara HHA, Faca VM, Candido Dos Reis FJ, et al. The contribution and perspectives of proteomics to uncover ovarian can-cer tumor markers. Transl Res 2019;206:71-90.
21.Feldreich T, Nowak C, Fall T, Carlsson AC, Carrero JJ, Ripsweden J, et al. Circu-lating proteins as predictors of cardiovascular mortality in end-stage renal disease. J Nephrol 2019;32:111-9.
22.Tasaki S, Suzuki K, Kassai Y, Takeshita M, Murota A, Kondo Y, et al. Multi-omics monitoring of drug response in rheumatoid arthritis in pursuit of molecular remis-sion. Nat Commun 2018;9:2755.
23.Toribio M, Fitch KV, Stone L, Zanni MV, Lo J, de Filippi C, et al. Assessing statin effects on cardiovascular pathways in HIV using a novel proteomics approach: Analysis of data from INTREPID, a randomized controlled trial. EBioMedicine 2018;35:58-66.
24.Assarsson E, Lundberg M, Holmquist G, Bjorkesten J, Thorsen SB, Ekman D, et al. Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, speci-ficity, and excellent scalability. PLoS One 2014;9:e95192.
25.Lundberg M, Eriksson A, Tran B, Assarsson E, Fredriksson S. Homogeneous antibody-based proximity extension assays provide sensitive and specific detection of low-abundant proteins in human blood. Nucleic Acids Res 2011;39:e102. 26.Warnatz K, Denz A, Drager R, Braun M, Groth C, Wolff-Vorbeck G, et al. Severe
deficiency of switched memory B cells (CD27(1)IgM(-)IgD(-)) in subgroups of
patients with common variable immunodeficiency: a new approach to classify a heterogeneous disease. Blood 2002;99:1544-51.
27.Babicki S, Arndt D, Marcu A, Liang Y, Grant JR, Maciejewski A, et al. Heat-mapper: web-enabled heat mapping for all. Nucleic Acids Res 2016;44: W147-53.
28.Chong J, Soufan O, Li C, Caraus I, Li S, Bourque G, et al. MetaboAnalyst 4.0: to-wards more transparent and integrative metabolomics analysis. Nucleic Acids Res 2018;46:W486-94.
29.Giovannetti A, Pierdominici M, Mazzetta F, Marziali M, Renzi C, Mileo AM, et al. Unravelling the complexity of T cell abnormalities in common variable immuno-deficiency. J Immunol 2007;178:3932-43.
30.Seidel MG, Kindle G, Gathmann B, Quinti I, Buckland M, van Montfrans J, et al. The European Society for Immunodeficiencies (ESID) registry working definitions for the clinical diagnosis of inborn errors of immunity. J Allergy Clin Immunol Pract 2019.
31.Hel Z, Huijbregts RP, Xu J, Nechvatalova J, Vlkova M, Litzman J. Altered serum cytokine signature in common variable immunodeficiency. J Clin Immunol 2014; 34:971-8.
32.Perreau M, Vigano S, Bellanger F, Pellaton C, Buss G, Comte D, et al. Exhaustion of bacteria-specific CD4 T cells and microbial translocation in common variable immunodeficiency disorders. J Exp Med 2014;211:2033-45.
33.Schreeder DM, Cannon JP, Wu J, Li R, Shakhmatov MA, Davis RS. Cutting edge: FcR-like 6 is an MHC class II receptor. J Immunol 2010;185:23-7.
34.Wilson TJ, Presti RM, Tassi I, Overton ET, Cella M, Colonna M. FcRL6, a new ITIM-bearing receptor on cytolytic cells, is broadly expressed by lymphocytes following HIV-1 infection. Blood 2007;109:3786-93.
35.Zhong H, Hu X, Janowski AB, Storch GA, Su L, Cao L, et al. Whole transcriptome profiling reveals major cell types in the cellular immune response against acute and chronic active Epstein-Barr virus infection. Sci Rep 2017;7:17775.
36.Lee SW, Croft M. 4-1BB as a therapeutic target for human disease. Adv Exp Med Biol 2009;647:120-9.
37.Buchan SL, Rogel A, Al-Shamkhani A. The immunobiology of CD27 and OX40 and their potential as targets for cancer immunotherapy. Blood 2018;131:39-48. 38.Wang S, Wang J, Kumar V, Karnell JL, Naiman B, Gross PS, et al. IL-21 drives
expansion and plasma cell differentiation of autoreactive CD11c(hi)T-bet(1) B cells in SLE. Nat Commun 2018;9:1758.
39.Zumaquero E, Stone SL, Scharer CD, Jenks SA, Nellore A, Mousseau B, et al. IFNgamma induces epigenetic programming of human T-bet(hi) B cells and pro-motes TLR7/8 and IL-21 induced differentiation. Elife 2019;8.
40.Yellin M, Paliienko I, Balanescu A, Ter-Vartanian S, Tseluyko V, Xu LA, et al. A phase II, randomized, double-blind, placebo-controlled study evaluating the effi-cacy and safety of MDX-1100, a fully human anti-CXCL10 monoclonal antibody, in combination with methotrexate in patients with rheumatoid arthritis. Arthritis Rheum 2012;64:1730-9.
41.Fliegauf M, Bryant VL, Frede N, Slade C, Woon ST, Lehnert K, et al. Haploinsuf-ficiency of the NF-kappaB1 subunit p50 in common variable immunodeHaploinsuf-ficiency. Am J Hum Genet 2015;97:389-403.
42.Klemann C, Camacho-Ordonez N, Yang L, Eskandarian Z, Rojas-Restrepo JL, Frede N, et al. Clinical and immunological phenotype of patients with primary im-munodeficiency due to damaging mutations in NFKB2. Front Immunol 2019;10: 297.
43.Ansel KM, Ngo VN, Hyman PL, Luther SA, Forster R, Sedgwick JD, et al. A chemokine-driven positive feedback loop organizes lymphoid follicles. Nature 2000;406:309-14.
44.Kroenke MA, Eto D, Locci M, Cho M, Davidson T, Haddad EK, et al. Bcl6 and Maf cooperate to instruct human follicular helper CD4 T cell differentiation. J Immunol 2012;188:3734-44.
45.Wang X, Cho B, Suzuki K, Xu Y, Green JA, An J, et al. Follicular dendritic cells help establish follicle identity and promote B cell retention in germinal centers. J Exp Med 2011;208:2497-510.
46.Havenar-Daughton C, Lindqvist M, Heit A, Wu JE, Reiss SM, Kendric K, et al. CXCL13 is a plasma biomarker of germinal center activity. Proc Natl Acad Sci U S A 2016;113:2702-7.
47.Romberg N, Le Coz C, Glauzy S, Schickel JN, Trofa M, Nolan BE, et al. Patients with common variable immunodeficiency with autoimmune cytopenias exhibit hy-perplastic yet inefficient germinal center responses. J Allergy Clin Immunol 2019; 143:258-65.
48.Hennig C, Ilginus C, Boztug K, Skokowa J, Marodi L, Szaflarska A, et al. High-content cytometry and transcriptomic biomarker profiling of human B-cell activa-tion. J Allergy Clin Immunol 2014;133:172-80.e1-10.
49.Jensen LE. Interleukin-36 cytokines may overcome microbial immune evasion strategies that inhibit interleukin-1 family signaling. Sci Signal 2017;10.
