Primary immunodeficiency‐related genes in neonatal intensive care unit patients with various genetic immune abnormalities: a multicentre study in China

Abstract Objectives The present phenotype‐based disease classification causes ambiguity in diagnosing and determining timely, effective treatment options for primary immunodeficiency (PID). In this study, we aimed to examine the characteristics of early‐onset PID and proposed a JAK‐STATopathy subgroup based on their molecular defects. Methods We reviewed 72 patients (< 100 days) retrospectively. These patients exhibited various immune‐related phenotypes and received a definitive molecular diagnosis by next‐generation sequencing (NGS)‐based tests. We evaluated the PID‐causing genes and clinical parameters. We assessed the genes that shared the JAK‐STAT signalling pathway. We also examined the potential high risks related to the 180‐day death rate. Results We identified PID disorders in 25 patients (34.72%, 25/72). The 180‐day mortality was 26.39% (19/72). Early onset of disease (cut‐off value of 3.5 days of age) was associated with a high 180‐day death rate (P = 0.009). Combined immunodeficiency with associated or syndromic features comprised the most common PID class (60.00%, 15/25). Patients who presented life‐threatening infections were most likely to exhibit PID (odds ratio [OR] = 2.864; 95% confidence interval [CI]: 1.047‐7.836). Twelve out of 72 patients shared JAK‐STAT pathway defects. Seven JAK‐STATopathy patients were categorised as PID. They were admitted to NICUs as immunological emergencies. Most of them experienced severe infections and thrombocytopenia, with 4 succumbing to an early death. Conclusions This study confirmed that NGS can be utilised as an aetiological diagnostic method of complex immune‐related conditions in early life. Through the classification of PID as pathway‐based subtypes, we see an opportunity to dissect the heterogeneity and to direct targeted therapies.


INTRODUCTION
With improvements in health care, genetic diseases have become the leading causes of infant mortality in neonatal intensive care units (NICUs). 1 Among the life-threatening genetic conditions in NICUs, immune abnormalities, mostly those predisposing to or causing severe infectious diseases, have frequently been unascertained. Primary immunodeficiency diseases (PIDs) 2-4 are a heterogeneous group of inborn errors of immunity and are mostly caused by monogenic germline mutations. Some of them are uniformly fatal without early definitive therapy within the first few months of life. Advances in nextgeneration sequencing (NGS) technologies and clinical care for critically ill infants have resulted in more forms of PID being reported to manifest in early life, the phenotypic spectrum associated with PIDs having evolved and more novel causal genes having been identified, so that infants who would have died in previous eras are now surviving with significant morbidity and mortality.
Currently, the categorisation of PID is based on the clinical, immunological and genetic characteristics of the disease. 5,6 However, there are often overlapping clinical manifestations of genetic and non-genetic causes, such as those resulting from compromised immunity, 7 during early postnatal life, which contribute to the difficulty in defining PID unequivocally and treating it individually. Therefore, the contemporary classification of human disease 8 was intrinsically limited for this subgroup. With the advent of genomewide molecular diagnostic techniques, identifying the underlying genetic defects has been widely used to classify diseases, predict disease progression and determine effective treatment options. [9][10][11][12] Some studies [13][14][15] indicated that mutations in different genes could affect the same cell signalling pathway and result in comparable clinical phenotypes; this finding has helped us understand genomic similarities and the difference between closely related clinical disorders. Accordingly, we hypothesised that paediatric patients who present immune-related conditions with suspected genetic aetiologies might possess monogenic mutations and might share biological mechanisms involving diseaseassociated genes from the same or related pathways. With this, we would connect diseases through their clinical phenotypes and associated genes and guide targeted therapeutic interventions. To date, studies investigating the roles of mammalian/mechanistic target of rapamycin (mTOR) 16 and nuclear factor-jB (NF-jB) 17 signalling pathways in immune dysregulation diseases have arguably a great effect in the inventories of mechanism-based therapies, such as infliximab, an anti-TNF agent in ameliorating autoinflammatory symptoms in patients with A20 haploinsufficiency, 18 and sirolimus, an mTOR inhibitor used to treat activated phosphoinositide 3-kinase d syndrome (APDS). 19 However, studies on the associations between genetic aberrations in specific pathways and immune deficiencies in critically ill infants are currently sparse, and their clinical implications for precision medicine are building.
In this study, we utilised NGS technology to recognise disease-causing mutations in PID in a selected cohort of unrelated critically ill infants presenting with various immune-related conditions in China. We perceived that a lifethreatening infection was the key clinical feature to identify a suspected PID in very early life. We also found that the genetic defects of some patients could be mapped to the Janus kinase (JAK)-signal transducer of activators of transcription (STAT) pathway and that this pathway was shared by some patients with severe PID disorders.

Patient characteristics
Seventy-two patients fulfilling the designated clinical inclusion criteria were identified as eligible for final analysis (Figure 1). Among these patients, 44 (61.11%, 44/72) were male. NGS method included panel analysis (n = 12, 16.67%) and exome sequencing (ES) (n = 60, 83.33%). A summary of all pathogenic variant types of 46 genes and nine copy number variations (CNVs) is provided in Supplementary table 1. Consanguinity was denied in 70 cases. Patients were born in nine different provinces in China and were unrelated. The mean AE standard deviation (SD) for birth weight was 3063.06 AE 746.05 g, and the percentage of premature neonates (< 37 weeks of gestation) was 25% (18/72). The clinical phenotypes were present at birth or in early age (0-97 days of age), with a median age of onset at 7.50 days. Although these infants had diverse clinical features, they exhibited one or more of the immune-related phenotypes at admission or during hospitalisation. In total, 48 patients were vaccinated with BCG and/or hepatitis B within 3 days of birth. No inoculation site infection and left axillary lymphadenectasis co-occurred with the immune-related phenotypes. Vaccinations for the remaining 24 patients were delayed because of neonatal onset and admissions to NICUs immediately after birth.

Genetic spectrum and clinical implications of PID
In total, 25 patients (34.72%, 25/72) were diagnosed with PID based on the molecular aetiology (Table 1). Thirteen different clinical diagnoses were made, which were separated into five groups according to the International Union of Immunological Societies (IUIS) classification. 20 Figure 2 showed the percentage of patients in each PID group according to the molecular diagnosis. Combined immunodeficiencies with associated or syndromic features were the most common PID (60%, 15/25), followed by diseases of immune dysregulation (20%, 5/25), immunodeficiencies affecting cellular and humoral immunity (8%, 2/ 25), congenital defects of phagocyte number or function (8%, 2/25) and defects in intrinsic and innate immunity (4%, 1/25).
Overall, NGS revealed 13 different PID disorders associated with 12 genes harbouring 18 diseasecausing variants and 1 CNV (Table 1, Figure 2). Six patients (P41, P39, P47, P64, P65 and P67) were diagnosed with ectodermal dysplasia with immunodeficiency (EDA-ID) because of IKBKG deficiency (OMIM: 300248). Four of them (P41, P47, P65 and P67) presented with skin lesions after birth and were treated with daily applications of emollients; all four patients achieved remission. The remaining two patients (P39 and P64) were initially suspected to have immune deficiency because they manifested severe and recurrent infections. Various antibiotics (ceftazidime, cefepime and meropenem) were applied, but they both died of blood culture (BC)positive sepsis during early infancy. P64 was positive for Acinetobacter baumannii, and P39 was positive for Klebsiella pneumonia. Four patients (P12, P13, P14 and P61) were diagnosed with IL-10Ra deficiency (OMIM: 146933) caused by various pathogenic variants in the IL10RA gene. These patients presented with intractable diarrhoea about 7 days after birth. Mouth ulcers manifested in 2 cases (P14 and P61), and extraintestinal manifestations included perianal abscesses (P13, P14 and P61) and perianal skin tags (P14). Four patients with Kabuki syndrome 1 (OMIM: 602113) (P52, P66, P70 and P71) had mutations in the KMT2D gene and demonstrated classic facial features and severe or recurrent infections during hospitalisation. Patient P52 presented with severe diarrhoea, pneumonia, cardiac anomalies and heart failure during the neonatal period; he underwent heart surgery at the age of 2 months, and his condition has since improved. NGS confirmed pathogenic variants in CHD7 in P48 and P72; both patients were diagnosed with CHARGE syndrome (OMIM: 608892) and presented with a combination of cleft palate, heart malformations and immune abnormalities. The remaining nine gene variants, including 1 CNV, were only identified in 1 infant per variant. These gene variants were characterised by various aforementioned immune abnormalities; among them, P17, who was diagnosed with leucocyte adhesion deficiency type 1 (LAD1) (OMIM: 600065), received haematopoietic stem cell transplantation (HSCT) and survived at 180 days of interview. P28 underwent heart surgery and was in clinical remission. Unfortunately, four patients (P1, P15, P59 and P69) with pathogenic variants in JAK3, IL2RG, SPINK5 and CYBB did not respond to the current treatments; their parents agreed to withdraw them from the treatments, and they passed away soon after.

Molecular and clinical features of non-PID patients
A molecular diagnosis of PID could not be made for the remaining 47 patients. Given the aforementioned immune abnormalities, further robust investigation is warranted to assess whether specific immune abnormalities were more likely to be associated with a diagnosis of   Table 2, and the identified gene distribution with different ages of onsets is shown in Figure 3a and b.
The JAK-STAT pathway-associated variants in JAK3, IL10RA, IL2RG and CXCR4 are loss-offunction mutations. In addition, seven patients with these defects can be categorised as PID according to the current criteria. Four cases (P12, P13, P14 and P61) had diseases of immune dysregulation, 2 (P1 and P15) had immunodeficiencies affecting cellular and humoral immunity, and 1 (P16) had defects in intrinsic and innate immunity. We found all of them consistently experienced severe infections (P1, P12, P13, P14, P15, P16 and P61). Four (P12, P13, P14 and P61) were diagnosed with IL-10Ra deficiency (OMIM: 146933) and treated with various antibiotics; it seemed that supportive treatments such as immunoglobulin transfusion, total parenteral nutrition, concentrated red blood cell transfusion and albumin transfusion to be more effective to treat. A female patient (P12) died at 26 days because of a serious systemic infection, which led to multiple organ failure. The remaining three patients (P13, P14 and P61) were in clinical remission: P13 was treated with thalidomide; P61 was transferred to a gastroenterologist who planned to provide HSCT for him; and P14 was given a highly hydrolysed formula.
Five patients with JAK-STATopathy (P2, P3, P4, P5 and P62) cannot be categorised as PID. Four patients who were correspondingly diagnosed with Costello syndrome (OMIM: 218040) (P2), Noonan syndrome 5 (OMIM: 611553) (P4) and Noonan syndrome 1 (OMIM: 163950) (P5 and P62) displayed recurrent infections and heart abnormalities constantly. Besides, both P2 and P4 had typical facial features (epicanthal folds, low set and prominent ears). P3 had no signs of infection, but presented with refractory   (Supplementary table 3). Laboratory tests revealed leucocyte counts of 1.60-46.37 9 10 9 L À1 (14.93 [9.19-22.63]; reference value: 4-10 9 10 9 L À1 ), a CRP level < 8 to 160 mg L À1 (8.  The remaining nine deceased  infants (P1, P15, P22, P23, P24, P30, P39, P63 and  P64) presented clinically between 28 and 100 DOL (Figure 3a and b). We found that early onset of disease was significantly associated with a high 180-day death rate (P = 0.009) (see Supplementary appendix 1 for details) (Supplementary table 4). The optimal cut-off value of the age of onset was set at 3.5 days of age (Supplementary figure 1). Eight patients (P1, P12, P15, P39, P46, P63, P64 and P69) diagnosed with PID died before 6 months of age (Table 1) (Figure 3a). The 180day mortality rate was 32.00% (8/25), and all died because of severe infections. They presented with the following symptoms indicating severe infections such as sepsis or sepsis shock or needing ventilation support or vasopressor therapy or presenting organ dysfunction. Two patients (P39 and P64) had positive results of bacterial cultures and no obvious responses to antimicrobial therapies. The statistical analysis revealed no significant association between the mortality and the diagnosis of PID. Moreover, no significant differences were observed when we compared the 180-day mortality rate with that of non-PID patients (Supplementary table 4).
Four patients (P1, P5, P12 and P15) (33.33%, 4/ 12) with genetic deficits in the JAK-STAT pathway died (Table 2): three died of severe infection without positive results of bacterial cultures and one died of bleeding from a low platelet count. We also found a non-significant association between the perturbations on this pathway and the 180-day mortality (Supplementary table 4).

DISCUSSION
Herein, we report the first multicentre study of a selected cohort of unrelated critically ill infants with various immune-related conditions in China. Among 72 patients with a definite molecular aetiology, approximately one-third had PID. Patients who presented with life-threatening infections were most likely to develop PID. Twelve patients may have shared similar biological mechanisms based on their underlying gene mapping to the same JAK-STAT pathway. Seven of them were classified with PID and were considered immunological emergencies, with four deceased during hospitalisation or soon after, indicating that this subgroup required targeted therapies rapidly. Finally, we found that the early onset of disease was significantly associated with a high 180-day mortality.
In this study, we reported 25 newly diagnosed infantile PID patients. Our diagnostic yield (34.72%, 25/72) was similar to that observed in European 21 and Egyptian patients. 22 In a recent systematic review, Yska et al. 23 reported that the diagnostic yield of NGS in mixed PID groups ranged from 15% to 46% (median = 25%), which is consistent with that derived from our cohort. When we inspected the associated clinical signs 24 among patients with PID from NICUs, we found that severe infections were the most common presentations, followed by atopic dermatitis and protracted diarrhoea. As for the immune phenotypes of PID patients, the results of seven patients (P1, P12, P13, P14, P15, P16 and P61) who completed routine immunologic investigations showed limited disease-associated features (Supplementary table 3). Age-dependent immunological memory and functional and quantitative deficiencies in antigen-presenting cells and phagocytes during early life 25 might explain the high immunological false-negative results in earlyonset PID. Still, in neonates, the inexperience of antigenic exposure, along with enhancement of tissue-protective immunosuppressive mechanisms, often results in homogeneous immune-related symptoms in critically ill patients with different disorders. 25 This is supported by our findings that 47 non-PID patients also manifested the aforementioned immune-related features during early life. Therefore, a timely application of NGSbased tests in these selected patients is necessary to identify or exclude monogenic PID. NGS-based tests are beneficial for the application of specific treatments and avoidance of nonspecific and ineffective drugs, thus resulting in an overall reduction in the healthcare costs for this subgroup. Notwithstanding, clinical signs are sometimes nonspecific or masked in the setting of critical conditions in early life. We noted that patients who presented with life-threatening infections were more likely to be diagnosed with PID; diseaseassociated hypogammaglobulinaemia was observed in two severe combined immunodeficiency (SCID) patients (P1 and P15) and one WHIM syndrome patient (P16). Thus, routine haematology and immunologic and microbiological tests may help exclude other causes of the immune-related conditions in early life and provide some clues for a PID aetiology. In addition, an estimated not-toolow potential incidence of SCID in China derived from our study (8.00%, 2/25) and that of Pilania et al. 26 indicates that T-cell receptor excision circle (TREC) analysis 27 is an accessible tool for SCID screening. These insights might potentially increase the diagnostic rate of PID and actualise personalised medicine for more patients.
Finally, we explored whether the defects of the JAK-STAT signalling pathway 28,29 encompass immune-related conditions. Our results confirmed that the 12 critically ill infants with genetic defects in JAK-STAT signalling consistently presented with immune abnormalities. Additionally, seven patients with defects in IL10RA, IL2RG, CXCR4 and JAK3 could be grouped into the PID group. High rates of severe infection (71.43%, 5/7) and thrombocytopenia (71.43%, 5/7) were observed in this group; antibiotics were rendered ineffective. These findings suggest that individuals, who develop a spectrum of similar symptoms, such as recurrent pneumonia, other infections, inflammatory bowel disease and thrombocytopenia, could be the key to determining the underlying genetic defects associated with the JAK-STAT pathway. The similarity between phenotypes manifesting in different monogenetic diseases in this study reflects the modular nature of related genes that share the same pathway. In addition, the identification of pathway-specific disease-causing variants has major therapeutic implications. From our results, SCID (OMIM: 600802), early-onset inflammatory bowel disease 28 (OMIM: 613148) and X-linked SCID (OMIM: 312863) should be considered as infantile emergencies. Unfortunately, patients diagnosed with these diseases in our cohort died as a result of severe infections or unsuitable management. However, other studies demonstrated that the survival of children with SCID undergoing cord blood transplantation (UCBT) 30 or HSCT 31 before 3.5 months of age HSCT has improved dramatically. Together, these findings reveal that early-onset PID disorders with aetiologic defects mapped to JAK-STAT signalling pathway should be considered immunological emergencies necessitating rapid and specific therapies.
The main strengths of this study include the recruitment of biological samples of 72 participants with early-onset immune-related conditions and a definitive molecular aetiology, which ensured satisfactory statistical power of the sample and limited the influence of confounding variables that studies with late-onset patients tend to present. The main limitation is the retrospective design of the study, which is prone to bias. Moreover, NGS diagnostics might fail to detect certain types of variants, such as balanced structure variation and deep intronic variants, which might be a source of minimal distortion. A prospective study with a larger sample size and thorough tests such as routine haematology and immunologic tests, whole-exome sequencing and/ or whole-genome sequencing in NICU settings would help further validate the utility of NGS as a first-tier approach in infants with complex immune-related conditions and provide more indepth insight into other sparse signal pathways.
In conclusion, a heterogeneous group of inborn errors of immunity present in infancy and result in significant morbidity and mortality. Therefore, a lower threshold is required for requesting genetic screening when a genetic aetiology is suspected in infants with various immune-related conditions during the neonatal period or shortly after it. Prompt pathway-specific categorisation for a group of diseases might be the first step to realise 'precision medicine' for NICU patients.

Patient selection algorithm
The patient selection algorithm is summarised in Figure 1. Patients admitted to NICUs at three tertiary hospitals in Shanghai, China: Xinhua Hospital (XH), Children's Hospital (CH) and Shanghai Children's Medical Center (SCMC), between January 2016 and December 2019, were reviewed. All those patients having undergone NGS were collected. Two NGS options were eligible: panel sequencing of the patient who presented specific clinical phenotypes and ES of the patient who presented with multiple congenital abnormalities or syndromic features or genetic conditions of no defined causes. The clinical outcome was the mortality (yes or no) obtained by telephone interview at the age of 6 months. Parents provided written informed consent for the testing itself and agreed to pay for the assay. The analysis and publication of data related to the study were approved by the Medical Ethical Committee of XH Hospital (Approval Number: XHEC-D-2020-164).
Inclusion criteria included the following: (1) less than 100 days, (2) recent presentation with one or more of the following immune-related conditions: ① increased susceptibility to life-threatening infections (the patients with life-threatening infections are eligible if they have sepsis or sepsis shock, 32,33 or if they need ventilation support or vasopressor therapy or present organ dysfunction, coupled with suspected infection); ② intractable diarrhoea; ③ atopic dermatitis; ④ failure to thrive secondary to recurrent infections or intractable diarrhoea; ⑤ leucocytosis/ neutropenia; ⑥ thrombocytopenia; ⑦ autoimmune haemolytic anaemia; ⑧ a positive family history of immunerelated diseases in first-degree or second-degree relatives, and (3) no definitive clinical diagnosis. The exclusion criteria were (1) patients with aneuploidy; (2) patients who were positive for HIV screening; (3) patients where a genetic diagnosis was made within 180 days of age after discharge other than the initial panel or ES; and (4) the families who declined the 6-months-of-age telephone interviews.
Each patient was evaluated for enrolment on a case-bycase basis after consultation between a clinical geneticist and a neonatologist. Overall, 143 patients were recruited to our data set, where 72 of them were confirmed as having a genetic aetiology. Consequently, 72 patients were included in the final analysis. Data regarding gestational age, birthweight, family history and related clinical and molecular results were extracted from the medical records. The phenotypes of the affected infants were further translated into Human Phenotype Ontology (HPO) 34 terms (see Supplementary appendix 1 for details).

Clinical indices
Laboratory evaluations were performed in the study as indicated, including complete blood counts, differential count and CRP. The basic immune components and function were not consistently available owing to the retrospective study design and date of diagnosis in some patients, so that we could not provide a complete immunological analysis for every patient and subgroup comparisons.

Genetic analysis
DNA was extracted from the peripheral whole blood of 143 patients and their parents, if available, using a QIAamp Blood DNA Mini Kit (Qiagen GmbH, Hilden, Germany). For ES, the capture probes were those used in GenCap Custom Exome Enrichment Kits (MyGenostics, Beijing, China) or TruSight Rapid Capture Kits (Illumina, Inc., San Diego, CA, USA). Thirteen specific disease panels were used (MyGenostics, Beijing, China) (see Supplementary appendix 1 for details). The resulting libraries were sequenced by a HiSeq 6000 platform (Illumina). The Burrows-Wheeler Aligner (BWA) (v.0.5.9-r16) was used to align the reads to the human reference genome (GRCh37/ hg19). CNVs and small variants were identified using VarScan 2 or Genome Analysis Toolkit (GATK) (4.0.10.1).
Potential CNVs identified by WES, if any, were further examined by karyotype testing or chromosomal microarray analysis (CMA). Deleterious variants detected by NGS were confirmed via Sanger sequencing. PolyPhen-2 and sorting intolerant from tolerant (SIFT) were used to predict the pathogenic effects of unreported variants. Pathogenicity was interpreted according to the five-tier classification system recommended by the American College of Medical Genetics and Genomics. [35][36][37] Statistical analysis Data were analysed using SPSS 23.0. Categorical variables are displayed as percentiles. The chi-square analysis was used to compare individual subgroup association with different clinical symptoms. A P-value < 0.05 was considered statistically significant.