Diagnostic Yield and Therapeutic Consequences of Targeted Next-Generation Sequencing in Sporadic Primary Immunodeficiency

Introduction: Primary immunodeficiencies (PIDs) are a heterogeneous group of disorders characterized by increased susceptibility to infections, immune dysregulation, and/or malignancy. Genetic studies, especially during the last decade, led to a better understanding of the pathogenesis of primary immunodeficiencies and contributed to their classification into distinct monogenic disorders falling under one of the >430 currently known inborn errors of immunity (IEI). The growing availability of molecular genetic testing resulted in the increasing identification of patients with IEI. Here, we evaluated the diagnostic yield and the clinical consequences of targeted next-generation sequencing (tNGS) in a cohort of 294 primary immunodeficiency patients, primarily consisting of cases with sporadic primary antibody deficiency. Method: We have custom designed a tNGS panel to sequence a cohort of PID patients. Agilent’s HaloPlex Target Enrichment System for Illumina was used for DNA target enrichment. Results: tNGS identified a definite or predicted pathogenic variant in 15.3% of patients. The highest diagnostic rate was observed among patients with combined immunodeficiency or immune dysregulation, for whom genetic diagnosis may affect therapeutic decision-making. Conclusion: Next-generation sequencing has changed diagnostic assignment and paved the way for targeted therapeutic intervention with agents directed at reverting the disease-causing molecular abnormality or its pathophysiological consequences. Therefore, such targeted therapies and identifying the genetic basis of PID can be essential for patients with manifested immune dysregulation as conventional immunomodulatory regimens may exert an immunosuppressive effect, aggravating their immunodeficiency or may only inadequately control autoimmune or lymphoproliferative manifestations.


Introduction
Primary immunodeficiencies (PIDs) or inborn errors of immunity (IEI) are heterogeneous disorders caused by a deficiency or a defect in one or more components of the immune system [1,2]. So far, pathogenic variants in >430 genes have been reported to cause PID. However, the strictly monogenic aetiology of PID seems questionable when considering the incomplete penetrance and variable expressivity of genetic variants reported as diseasecausing [3]. Variable expressivity, including variable age of onset and a broad spectrum of manifestations, ranging from "infections-only" to disease complicated with immune dysregulation and/or malignancy, makes a timely diagnosis of PID difficult, which leads to an increased rate of morbidity and mortality [4]. The clinical diagnosis of PID most commonly falls under one of the primary antibody deficiencies (PADs), especially common variable immunodeficiency (CVID) [5,6]. In the last decade, the increasing availability of next-generation sequencing (NGS) technologies led to identifying the involvement of new genes in PADs. It aided the characterization of the molecular pathways involved in human B cell development and function [2, 7,8]. The most common symptomatic PAD, CVID [7], is still considered a largely polygenic disorder, though according to more recent reports, the proportion of monogenic forms has increased, exceeding 20% of cases [8][9][10].
Genetic testing is currently routinely performed in most medical centres treating PID patients, resulting in an increasing number of monogenic IEI. Consequently, several studies of cohorts of patients with the same IEI that evaluated their phenotypes and outcomes have been published over the last couple of years, for example, identifying an increased risk of malignancies or association of particular IEI with systemic rheumatic or interstitial lung disease may alter diagnostic and follow-up procedures [11][12][13]. Further, in case of particular monogenic defects, individualized therapies with agents tailored to revert consequences of the disease-underlying genetic defect can be considered. For example, JAK-inhibitors may help treat immune dysregulation in the context of STAT3 or STAT1 gain-of-function [14], PI3Kδ-specific inhibition is effective against lymphoproliferative manifestation of the activated PI3Kδ syndrome [15], and abatacept may be effective in treating enteropathy and autoimmunity in patients with CTLA4 insufficiency or LRBA deficiency [16,17].
Overall, timely molecular diagnosis and setting-specific decisions may reduce the harms associated with PIDs and favour a better disease outcome. Hence, we employed a targeted NGS (tNGS) approach to identify the genetic background of immunodeficiency in a cohort of 294 patients, most of whom were sporadic cases of a PAD, aiming at increasing the diagnostic yield and optimizing patients' medical care.

Study Cohort
This single-centre study included a total of 294 patients with a PID, visiting the immunology outpatient clinics of either the Department of Paediatrics (N = 28) or the Department of Rheumatology and Immunology (N = 266) of the Hannover Medical School. The patients were clinically diagnosed with PID according to the European Society for immunodeficiencies criteria [18]. All relevant clinical and immunological data were obtained from the patient's medical files. Those included serum immunoglobulins, peripheral lymphocyte immunophenotyping, patient's clinical history of infections, bronchiectasis (computed tomographyconfirmed), autoimmune cytopenias, such as autoimmune haemolytic anaemia, idiopathic thrombocytopenic purpura, organspecific autoimmunity (including vitiligo, psoriasis, insulin-dependent diabetes mellitus, autoimmune thyroiditis, primary thyroid failure, atrophic gastritis, and rheumatic disease including rheumatoid arthritis, Sjögren's syndrome, systemic lupus erythematosus, and seronegative arthritis diagnosed according to the American College of Rheumatology [ACR]/European League Against Rheumatism [EULAR] classification criteria), benign lymphoproliferation, granulomatous disease, enteropathy, and malignancies. Benign lymphoproliferation was defined as splenomegaly (≥11 cm on palpation or ultrasound, including the previous splenectomy of an enlarged spleen) and/or persistent lymphadenopathy (on palpation, ultrasound, computed tomography, or magnetic resonance scan). Granulomatous disease was defined as at least 1 biopsy-proven unexplained granuloma, excluding Crohn's disease-associated granulomas. Enteropathy included all cases of biopsy-proven inflammatory bowel disease (ulcerative colitis and Crohn's disease) and intestinal hyperlymphocytosis (lymphocytic infiltration of the interepithelial mucous, the lamina propria and/or the submucosa). Malignancies included haematological and all other forms of cancer. All patients and their parents, in the case of paediatric patients, signed an informed consent form.

Targeted NGS
A customized panel of genes associated with PID (online suppl.  [19]. In brief, DNA was fragmented using a restriction enzyme master mix prepared following the kit's protocol, and digestion was validated by gel electrophoresis. DNA fragments were hybridized to the HaloPlex probe capture library by adding the Hybridization Master Mix and Indexing Primer Cassettes. After an incubation step, the hybridized DNA fragments were captured using HaloPlex Magnetic Beads, that is, a biotin-streptavidin system and washed. In order to create circular fragments, the ends were ligated by adding the kit's ligation master mix. Subsequently, the target libraries were amplified by polymerase chain reaction. Finally, the amplified target libraries were purified with the help of AMPure XP beads and washed in ethanol. Target enrichment was validated using an Agilent TapeStation. Samples were pooled in equimolar amounts. The libraries were subjected to denaturation by adding NaOH and diluted to a final concentration of 8-12 pM. Sequencing was performed on an Illumina MiSeq system using an Illumina v2 reagent kit, following the manufacturer's protocol. Data analysis was performed with the help of Agilent's SureCall software. Most likely disease-causing candidates were confirmed by Sanger sequencing. Familial segregation was examined when samples were available.

Variants Filtering Procedure
FastQ files were aligned to the human reference genome (UCSC hg19, GRCh37) and analysed using Agilent Technologies -Sure-Call NGS software. Variants were selected according to criteria at the variant level, including allele frequency (AF), variant annotation, and potential functional effect. Using databases of variants (e.g., dbSNP, 1000 Genomes Project, Exome Aggregation Consortium, and Genome Aggregation Database) and disease-causing variants (HGMD and OMIM), we selected all rare or private variants with an AF of <1%. Furthermore, we kept nonsense variants, variants affecting the splice site, frameshift, in-frame indels, start or stop codon changes, as well as missense variants that were predicted to be deleterious by having a combined annotation-dependent depletion (CADD) score >15 and a mutation significance cutoff score below the CADD score. Also, variants were tested by Mu-tationTaster [20], Provean, and Human Splicing Finder [21] for the pathogenicity.

Statistical Analysis
For statistical calculation, we used GraphPad prism 5.00 (GraphPad, La Jolla, San Diego, CA, USA). Descriptive statistics are reported as the median and interquartile range in continuous variables and as counts and percentages for dichotomous variables. The χ 2 test compared categorical variables. Non-categorical variables were compared with the Mann-Whitney test. All comparisons were 2-tailed, and p < 0.05 was considered significant.

Clinical Characterization of PID Patients
Patients' demographic data and characteristics, including the clinical diagnosis and manifestations of PID, are summarized in Table 1. Most of the 294 PID patients who underwent tNGS displayed a PAD (263/294, 89.5%), more commonly diagnosed as CVID (227/294, 77.2%).

Factors Affecting the Diagnostic Yield of tNGS
Overall, genetic testing utilizing tNGS identified a likely genetic diagnosis in 15.3% of studied patients. The diagnostic yield of tNGS could be different in distinct patient subgroups. We therefore evaluated the effect of fac-  Previous report on the involvement of the mutated gene in primary immunodeficiency. tors such as consanguinity, family history of PID, age at diagnosis of PID, the clinical diagnosis of PID, and its clinical manifestations on the diagnostic rate of tNGS. As expected, consanguinity or familial history of PID led to a considerably higher diagnostic rate (Table 3). In particular, in the subgroup of patients with consanguinity or a family history of PID (37/294), the diagnostic performance of tNGS would be 48.6% (18/37). In contrast, excluding those patients would result in reduction of the diagnostic yield to 10.5% (27/257). Regarding the clinical diagnosis of PID, the highest diagnostic rate has been identified in the subgroup of patients with CID. Among evaluated clinical manifestations, a predicted diseasecausing variant was more commonly found among the subgroup of patients with enteropathy and organ-specific autoimmunity.

Therapeutic Consequences of Genetic Diagnosis through tNGS
Genetic testing through tNGS had a higher diagnostic rate among patients with immune dysregulation, manifesting as organ-specific autoimmunity or enteropathy, identifying among else, monoallelic gain-of-function variants in STAT1 and STAT3, CTLA4 variants leading to CTLA4 insufficiency, PIK3R1 variants, likely resulting in hyperactivation of PI3Kδ, and biallelic LRBA variants, likely causing LRBA deficiency. As discussed above, targeted therapies can be considered for those genetic diagnoses, especially for treating manifestations of immune dysregulation, such as autoimmune and/or lymphoproliferative disease.
Based on genetic diagnosis, already 2 patients diagnosed with STAT3 gain-of-function are treated with baricitinib, due to their progressive interstitial lung disease (P.227, P.265). A paediatric patient harbouring a STAT3 gain-of-function variant (P.278) developed severe ILD, which led to aggressive immunosuppressive treatment with glucocorticosteroids and hydroxychloroquine mycophenolate mofetil, intravenous immunoglobulins, and etanercept, which were all ineffective in controlling its progressive course. Consequently, this child underwent bilateral lung transplantation at the age of 14 years and died 12 months later due to treatment-resistant chronic allograft dysfunction. The genetic analysis by tNGS has been only performed post-mortem and revealed a mutation in STAT3 (c.1276T>C, p.C426R), whose gain-of-function effect has been confirmed [38]. Timely genetic diagnosis, in that case, would have led to employing treatment with a JAKinhibitor or tocilizumab, which appears more effective in treating immune dysregulation in the context of STAT3 gain-of-function syndrome. Similar to the STAT3 gain-offunction patients, a patient with STAT1 gain-of-function (P.101) displayed steroid-refractory autoimmune haemolytic anaemia and lymphopenia, which were successfully controlled after introducing treatment with baricitinib. Among identified patients with a CTLA4 mutation, 2 displayed a therapy-refractory enteropathy (P.165, P.215) and will be offered an abatacept treatment.
Besides targeted treatment of immune dysregulation, genetic diagnosis may lead to timely haematopoietic stem cell transplantation (HSCT), especially in patients with CIDs. Among tested patients, we identified a patient with RAG1 deficiency (P.64). This patient was diagnosed with progressive multifocal leukoencephalopathy, which had a lethal outcome. Considering the link between RAG1 deficiency and SCID or CID [52,53], earlier genetic diagnosis in that patient could have led to a timely consideration of HSCT. Similar is the case of the patient diagnosed with IκBα gain-of-function, who currently suffers from recurrent infections, including pneumonia, and already has severe bilateral bronchiectasis (P.236) [34]. Given the lethal outcome of this IEI, timely genetic diagnosis at the age of In summary, in 4/294 (1.4%) patients, genetic diagnosis led to the introduction of a targeted therapy, tailored to revert the consequences of the identified genetic defect. However, considering patients with retrospective, postmortem diagnosis, or currently treatment-refractory immune dysregulation, the fact that some of the identified predicted pathogenic variants, such as those in LRBA or PIK3R1, have not been functionally validated, genetic diagnosis through tNGS could affect the follow-up and/or result in therapeutic consequence in substantially higher number of tested patients.

Discussion/Conclusion
tNGS has been employed in the present study to investigate the genetic basis of immunodeficiency in a cohort of patients who mainly had sporadic PID, falling under a PAD. This sequencing approach revealed a likely pathogenic genetic variant in approximately 15% of tested pa-tients. This diagnostic rate was relatively low, though similar to the one reported in some previous genetic studies, employing WES or WGS on cohorts of patients with sporadic immunodeficiency [9,50,59]. The diagnostic rate of tNGS was significantly higher in the subgroup of patients with a familial history of PID or those with an early onset of disease. CID or immunodeficiency complicated with enteropathy or organ-specific autoimmunity were additional factors associating with a significantly higher diagnostic rate. The overall low diagnostic rate of tNGS in the present study may be explained through the dependence of the diagnostic performance of genetic testing and tNGS on factors such as consanguinity or family history with PID [60], which were present in a minority of tested patients. Further, the limited diagnostic yield of tNGS may reflect the polygenic origin of PID and the pathogenic role of additional yet unidentified genetic and/or epigenetic modifiers [3]. Given the increasing number of genes linked to IEI, additional disease-causing genes may be identified in the near future. For this reason and as the gene panel tested in the present study did not include all currently known PID genes, employing a CI, confidence interval; CID, combined immunodeficiency; CVID, common variable immunodeficiency; ILD, interstitial lung disease; ns, not significant; OR, odds ratio; PID, primary immunodeficiency disorder; SPAD, specific antibody deficiency. + p < 0.05. ** p < 0.01. *** p < 0.001. 1 Including 3 patients from the same family. 2 Analysis based on 276/294 studied patients with the known year of diagnosis. 3A utoimmune hemolytic anemia (AIHA) and/or immune thrombocytopenic purpura (ITP). 4A topic dermatitis and/or allergic rhinitis and/or asthma. DOI: 10.1159/000519199 broader sequencing approach, such as WES, may have resulted in a higher diagnostic rate.
Overall, 3.1% of tested PID patients harboured a predicted pathogenic NFKB1 variant, which was the most common genetic defect among tested patients. This finding is in line with previous studies [61,62], which identified heterozygous loss-of-function variants in NFKB1 as the most common genetic defect in patients with CVID. In compliance with previous studies of large cohorts of PID patients with NFKB1 mutations, this subset of patients displayed variable immunodeficiency, ranging from infections-only antibody deficiency to CID and marked immune dysregulation [19]. Characterization of phenotypes of patients with immunodeficiency linked to heterozygous NFKB1 mutations revealed a relatively high prevalence of autoimmune and lymphoproliferative manifestations [19,32,61]. Together with the recent identification of NFKB1 mutations in patients with sheer rheumatic disease and secondary hypogammaglobulinaemia [63], the latter leads to the assumption that NF-κB1-related disease is primarily a condition of immune dysregulation rather than mere immunodeficiency [32]. The observation of progressive hypogammaglobulinaemia in currently asymptomatic mutation carriers and the likely concurrent progressive course of immunodeficiency [61] may lead to a later disease onset in some currently asymptomatic mutation carriers; immunodeficiency might be precipitated from the introduction of immunomodulatory regimens for their autoimmune manifestations [63]. The aforementioned finding emphasizes the importance of early molecular diagnosis of NF-κB1-related disease and the need to develop targeted NF-κB pathway-based therapeutic intervention.
The increasing availability of sequencing technologies, especially of NGS, has improved our understanding of the pathogenesis of PID and paved the way for targeted therapeutic intervention with agents directed at reverting the disease-causing molecular abnormality or its pathophysiological consequences [64,65]. Therefore, such targeted therapies and identifying the genetic basis of PID can be essential for patients with manifest immune dysregulation as conventional immunomodulatory regimens may exert an immunosuppressive effect, aggravating their immunodeficiency or may only inadequately control autoimmune or lymphoproliferative manifestations. Already employed individualized therapeutic approaches in the field of IEI include mTOR-or PI3Kδ-specific inhibitors in patients with mutations in PIK3CD, PTEN, or PIK3R1 causing activated PI3Kδ syndrome [29,49]; abatacept (CTLA4-Ig) in patients with LRBA deficiency; CTLA4 insufficiency or DEF6 deficiency [16,17,66]; jakinibs in patient with STAT1 or STAT3 gain-of-function [14]; and the CXCR4 antagonist plerixafor in patients with warts, hypogammaglobulinaemia, immunodeficiency, and myelokathexis syndrome [67]. Despite the lack of randomized controlled clinical trials, evidence from case reports and case series suggests both the excellent tolerability and efficacy of such targeted therapies. Together with their focused biological effect, it makes them attractive, especially when compared with conventional immunomodulatory regimens. Here, molecular diagnosis through tNGS resulted in treatment with baricitinib in a subgroup of patients with gain-offunction mutations in STAT1 or STAT3. Further, postmortem identification of therapeutically relevant mutations in 3 studied patients, who directly or indirectly died of treatment-refractory immune dysregulation, highlights the importance of early genetic diagnosis, especially in patients with the autoimmune or lymphoproliferative disease as this would have resulted in timely consideration of targeted therapy. In the current study, the rate of patients who received targeted therapies based on genetic diagnosis was low and may underestimate the clinical significance of genetic diagnosis. The main reasons for the lack of a therapeutic consequence were delayed genetic testing, including cases with post-mortem diagnosis, and the lack of functional validation of some of the identified variants. Further, genetic diagnosis should lead to genetic counselling of family members and family cascade testing, which can lead to genetic diagnosis in additional patients and may expand the clinical consequences of genetic testing.
Generally, in comparison to whole-exome (WES) or whole-genome sequencing (WGS), the targeted sequencing gene panel has an increased sequencing depth and allows rapid, cost-effective, and simultaneous sequencing of multiplexed samples in 1 sequencing run [68]. However, given the growing number of genes linked to IEI, it may require >1 gene panel for the genetic work-up PID. Despite testing for variations in genes most commonly linked to IEI or currently considered therapeutically relevant, the latter limitation of tNGS may underestimate the prevalence of monogenic disease in the studied cohort as discussed above. For this purpose and based on a recommendation of the IUIS experts committee, the best way to identify a disease-associated genetic variation in a new PID patient might consist of a tNGS followed by WES or WGS for unresolved cases [43]. However, given the currently dropping prices of WES and WGS, this suggestion may need to be revised. An additional limitation of our study is the in silico evaluation of the pathogenicity of identified variants, which may overestimate the prevalence of PID-related variants. Even in the case of a definite molecular diagnosis, for most patients harbouring an IEI-causing mutation, the disease does not follow Mendelian inheritance. The identified pathogenic variant is commonly associated with incomplete penetrance variable expressivity, rendering disease course and therapeutic choice difficult. Therefore, precision medicine, whose scope is to be served through routine genetic testing, may be further facilitated by identifying an additional germline or somatic genetic modification as well as epigenetic factors and identifying the role of environmental factors [69][70][71]. A better understanding of those factors could provide additional insight into disease pathogenesis's fundamental processes, offering a better perspective towards genetic counselling and precision medicine.
In conclusion, the employed Haloplex tNGS-approach is a time-and cost-effective sequencing method, providing a unique diagnostic tool for identifying known or novel variants in PID-associated genes. This approach had a high diagnostic rate, especially within the subgroup of patients with CID or immune dysregulation, for whom genetic diagnosis may affect therapeutic decision-making. However, given the increasing number of genes associated with IEI, which at the moment amount to >430, a panel-sequencing approach is ineffective in integrating all disease-relevant genetic loci, highlighting the importance of sequential sequencing, where negative-targeted NGS may be followed by WES or WGS. The latter may be especially relevant for patients who, in addition to antibody deficiency, display cellular immunodeficiency or those with complications such as autoimmunity, lymphoproliferation, or malignant disease as identifying the genetic cause of disease may affect follow-up procedures or therapeutic decision-making.