Introduction

Systemic lupus erythematosus (SLE) is a chronic autoimmune disease, characterized by the presence of autoantibodies and multisystem involvement. Pediatric SLE accounts for about 15% of all SLE patients. The phenotype may be different from adult-onset SLE, with more complex clinical manifestations, rapid progression, and poor prognosis. SLE has a multifactorial pathogenesis; however, the underlying etiology contributions of SLE including genetic, environmental, and immunologic factors have remained elusive.

Monogenic lupus is defined as SLE patients with either dominantly or recessively inherited pathogenic variants in a single gene with high penetrance [1]. Despite being rare, it has made a significant contribution to revealing the molecular pathogenesis of SLE. Early onset, familial, and syndromic SLE are the clinical features of this disease [2]. Up to 100 susceptibility loci for polygenic SLE, as well as around 50 monogenic causes of SLE or lupus-like phenotype, have been described to date [1,2,3,4,5,6,7,8,9]. These include pathways in complement deficiencies, type I interferon (IFN) signaling, RAS, and self-tolerance, among others (Fig. 1, as of March 2023) [1, 10]. The vast majority of genetic defects leading to monogenic lupus belong to complement and type I IFN pathways. Considering the complexity of lupus, the recognition of monogenic causes is becoming increasingly important as treatment options are developed based on an understanding of causal molecular pathways. However, it is still difficult for clinicians to identify monogenic lupus at an early stage so effective treatment can be initiated as soon as possible.

Fig. 1
figure 1

Pathways of genes associated with monogenic lupus. IKZF1 Ikaros family zinc finger proteins 1, P2RY8 P2Y receptor family member 8, PIK3CD phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit delta, RPKCD protein kinase C delta, TNFSF13B tumor necrosis factor superfamily member 13b, RAG recombination-activating gene, KRAS Kirsten rat sarcoma viral oncogene homolog, PTPN11 protein tyrosine phosphatase non-receptor type 11, SOS1 son of sevenless homolog 1, NRAS neuroblastoma rat sarcoma viral oncogene homolog, TLR7 toll-like receptor 7, TNFAIP3 tumor necrosis factor alpha-induced protein 3, PLCG1 phospholipase Cgamma 1, SBDS Shwachman-Bodian-Diamond syndrome, PTEN phosphatase and tensin homolog deleted on chromosome ten, SLC7A7 solute carrier family 7 member 7, FAS Fas cell surface death receptor, FASL Fas ligand, PEPD peptidase D, NEIL3 nei endonuclease VIII-like 3, CYBB cytochrome b-245, beta polypeptide, SAT1 spermidine/spermine N1-acetyltransferase 1, ACP5 tartrate-resistant acid phosphatase, ADA2 adenosine deaminase 2, ADAR1 adenosine deaminase RNA specific 1, DNASE1/2 deoxyribonuclease 1/2, DNASE1L3 deoxyribonuclease 1 like 3, DDX58 RNA sensor RIG-I, IFIH1 interferon induced with helicase C domain 1, ISG15 interferon-stimulated gene product 15, USP18 ubiquitin-specific peptidase 18, PSMA3 proteasome 20S subunit alpha 3, PSMB4 proteasome 20S subunit beta 4, RNASEH2A ribonuclease H2 subunit A, RNASEH2B ribonuclease H2 subunit B, RNASEH2C ribonuclease H2 subunit C, SAMHD1 SAM-domain- and HD-domain-containing protein 1, TMEM173 transmembrane protein 173, TREX1 three prime repair exonuclease 1, OTUD1 OTU deubiquitinase 1, PKB protein kinase B, NF-ĸB nuclear factor-kappa B

Full size image

The least absolute shrinkage and selection operator (LASSO), as a supervised learning method, is capable of constructing models using labeled data and predicting outcomes for unknown data. A nomogram chart is a convenient and intuitive risk assessment tool that integrates multiple factors. By considering various variables and incorporating the model, it aids healthcare professionals in swiftly and intuitively evaluating a patient’s risk.

We aim to summarize the clinical characteristics of monogenic lupus and develop an early clinical identification model applicable for assessing disease risk in pediatric patients, thus reducing the delay in diagnosing and treating monogenic lupus.

Methods

Study design and patients

This study was a retrospective case–control study. Inclusion criteria for the case group were as follows: (1) children under the age of 18 years, treated at Peking Union Medical College Hospital between June 2012 and December 2022, diagnosed with monogenic lupus; (2) confirmed to have a single pathogenic gene variant through genetic testing, with clinical presentations consistent with SLE classification criteria or exhibiting a lupus-like phenotype. Inclusion criteria for the control group were children with classic SLE who were admitted to our hospital within one year before or after the first admission date of the case group, matched in a 1:2 ratio. This study obtained approval from the Ethics Committee of Peking Union Medical College Hospital.

Definition

The diagnosis criteria for SLE are the 1997 American College of Rheumatology classification criteria or the 2012 Systemic Lupus International Collaborating Clinics classification criteria [11, 12]. A lupus-like phenotype is defined as the presence of involvement of two or more systems in the SLE-affected systems, as well as positivity for one or more autoantibodies, but does not meet the classification criteria for SLE.

Statistical analysis

Data analysis was conducted using R (version 4.1.1) and GraphPad Prism (version 9.0). Categorical variables were expressed as counts (%). Continuous variables were presented as mean ± standard deviation. A P value less than 0.05 was considered statistically significant.

Results

Demographic data

We included a total of 41 patients with monogenic lupus and 82 patients with classic lupus. Within the classic SLE group, the average age was 11.65 ± 1.87 years, with nine male patients (male:female ratio of 1:8.1), and six (7.3%) patients had a family history. In the monogenic lupus group, the average age was 3.47 ± 4.18 years, with 20 male patients (male:female ratio of 1:1.05), and nine (22%) patients had a family history. Both groups showed statistically significant differences in terms of age, gender, and family history (P < 0.05) (Table 1).

Table 1 Demographics of monogenic lupus in 41 patients
Full size table

Gene data

A total of 18 different gene mutations were identified among the 41 pediatric patients, as shown in Table 2. Out of these, 26 cases were associated with type I IFN pathway disorders, including six cases with adenosine deaminase 2 (ADA2) gene mutations, five cases with helicase C domain 1 gene mutations, four cases with three prime repair exonuclease 1 gene mutations, three cases with stimulator of interferon gene 1 mutations, two cases with ribonuclease H2 subunit C gene mutations, two cases with tartrate-resistant acid phosphatase gene mutations, two cases with proteasome subunit beta type-8 gene mutations, one case with RNA-editing enzyme adenosine deaminase RNA specific 1 gene mutation, and one case with sterile alpha motif and HD domain-containing protein 1 gene mutation. Additionally, five cases were linked to RAS-associated autoimmune leukoproliferative disorder, comprising two cases with Kirsten rat sarcoma viral oncogene homolog (KRAS) gene mutations and three cases with neuroblastoma RAS viral oncogene homolog gene mutations. Furthermore, four cases were related to immune tolerance pathways, with three cases having phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit delta gene mutations, and one case with an Ikaros family zinc finger proteins 1 gene mutation. There was also one case with C1s deficiency. Other mutated genes included two cases with tumor necrosis factor alpha-induced protein 3 gene mutations, one case with solute carrier family 7 member 7 gene mutation, one case with peptidase D gene mutation, and one case with Shwachman–Bodian–Diamond syndrome (SBDS) gene mutation. Some of these gene loci have been previously reported by our team [13,14,15,16], while others have been documented in various literature [17,18,19,20,21,22,23,24,25,26,27,28,29]. Nine loci from seven genes have not been reported (Table 2). Although two loci of KRAS have been reported, they were not related to lupus. In addition, our center reported for the first time that the SBDS gene was related to SLE [29].

Table 2 Forms of monogenic lupus in 41 patients
Full size table

Clinical characteristics

The primary clinical manifestations of both groups are compared in Table 3. Although skin and mucosal involvement were common in both groups, classic SLE was more likely to present with acute/subacute rashes (67.1%), primarily typical malar rashes (58.5%). In contrast, patients in the monogenic lupus group were more likely to exhibit chronic rashes (39%), particularly chilblain-like rashes (19.5%), nodular erythema (12.2%), and livedo reticularis (19.5%). Hematologic system involvement was more common in classic SLE, mainly manifested as hemolytic anemia (53.7%), whereas monogenic lupus primarily exhibited abnormal liver function (65.9%) in the gastrointestinal system. Neuropsychiatric involvement, on the other hand, was more common in monogenic lupus (46.3%), with abnormal muscle tone being the most common (22%), followed by seizure episodes (14.6%). Monogenic lupus also exhibited more significant respiratory system involvement (53.7%), primarily with pulmonary interstitial involvement (27.8%). Renal involvement was more frequent in the classic lupus group (56.1%). In terms of other clinical presentations, patients with monogenic lupus were more likely to have a history of recurrent infections (31.7%), accompanied by fever (63.4%), lymphadenopathy/hepatosplenomegaly (56.1%), and growth and developmental delay (48.8%). Regarding complications, the classic SLE group was more prone to cytomegalovirus (CMV) infection (70.7%). There were no statistically significant differences in other affected systems.

Table 3 Clinical characteristics of monogenic lupus in 41 patients
Full size table

While the monogenic lupus group can also exhibit various positive autoantibodies, the classic SLE group is more likely to present with lupus-specific antibodies, such as antinuclear antibodies, anti-double strand-DNA antibodies, anti-Smith antibodies, anti-ribonuclear protein (RNP) antibodies, anti-Sjogren's syndrome antigen A antibodies, anti-ribosomal RNP antibodies, anti-Ro-52 antibodies, anti-complement antibodies, anti-nucleosome antibodies, and lupus anticoagulants. Individuals in the classic SLE group are also more likely to have decreased complement levels, all of which demonstrate significant statistical differences, as shown in Table 4.

Table 4 Auxiliary inspection of monogenic lupus in 41 patients
Full size table

Head imaging findings of the two groups of patients, including head computed tomography, head magnetic resonance imaging/magnetic resonance angiography/magnetic resonance venography, are compared in Table 4. The monogenic lupus group showed a higher proportion of patients with intracranial calcifications and brain atrophy, with significant statistical differences.

Predictor selection

The variables that showed differences between the two groups in the univariate analysis were included in the LASSO regression analysis. After LASSO regression selection (Fig. 2a), the subsequent seven variables emerged as significant predictive factors for monogenic lupus: age of onset, history of recurrent infections, intracranial calcifications, growth and developmental delay, abnormal muscle tone, lymphadenopathy/hepatosplenomegaly, and chilblain-like skin rash. Notably, the variables in the model are independent. The regression coefficients for these variables were − 1.5864, 0.9920, 0.6654, 0.3318, 0.2103, 0.1179, and 0.0896, respectively (Fig. 2b). When the coefficients are scaled back to the original units of the variables, the resulting linear prediction model for monogenic lupus is as follows: 0.2906977 − 1.4124018 × age of onset + 15.2761768 × recurrent infections + 7.5899250 × intracranial calcifications + 3.5908622 × growth and developmental delay + 3.0491924 × abnormal muscle tone + 1.1751209 × lymphadenopathy/hepatosplenomegaly + 1.2324701 × chilblain-like skin rash.

Fig. 2
figure 2

LASSO regression coefficient profiles and model variables with their coefficients. a LASSO coefficient profiles of the radiomic featuresl; b model variables and their coefficients. LASSO least absolute shrinkage and selection operator, AUC area under the receiver operating characteristic curve

Full size image

Model validation and nomogram construction

The LASSO model achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.97 [95% confidence interval (CI) = 0.92–0.97] in the test set (Fig. 3a). We also constructed a logistic regression model and evaluated its performance, resulting in an AUC of 0.87 (95% CI = 0.75–0.87). Comparison between the two models demonstrated the superiority of the LASSO model (Fig. 3a). Furthermore, the LASSO model exhibited an accuracy of 0.86, a precision of 0.81, a recall of 1, and an F1 score of 0.89, indicating excellent predictive performance of this model. We assessed the risk score prediction for monogenic lupus using the ROC curve, which yielded an AUC of 0.98 (95% CI = 0.97–1.00) (Fig. 3b). Utilizing the Youden index, we identified a cutoff score of − 9.032299, suggesting that patients with SLE/lupus-like who have a predictive model score greater than − 9.032299 are at a higher risk of having monogenic lupus, with a risk probability of 0.766 (sensitivity = 92.7%, specificity = 98.8%). We have represented the risk factors in a nomogram (Fig. 4) to facilitate clinicians in intuitively assessing a patient’s risk of developing monogenic lupus.

Fig. 3
figure 3

Comparison and risk prediction of monogenic lupus using ROC analysis. a The comparison of ROC curves between LASSO regression and logistic regression; b ROC curve for risk score prediction of monogenic lupus. LASSO least absolute shrinkage and selection operator, ROC receiver operating characteristic, AUC area under the ROC curve

Full size image
Fig. 4
figure 4

Nomogram for probability of monogenic lupus

Full size image

Discussion

The understanding of the genetic mechanism of SLE in children has rapidly improved due to the emergence of monogenic lupus. The discovery of new or rare pathogenic gene mutations holds great significance in revealing the molecular pathways involved in the pathogenesis of the disease. Furthermore, such discoveries have the potential to facilitate the provision of individualized treatment. However, because the clinical phenotype of monogenic SLE is extensive and overlaps with that of classical SLE, it causes a delay in diagnosis and treatment. Currently, most reports on monogenic lupus are case studies, and there is a lack of early identification models for clinical practitioners to provide early clues for recognition. Therefore, this study aims to summarize the clinical characteristics of monogenic lupus and establish a clinical early identification model for it. Using a nomogram allows for rapid scoring of patients, indicating the probability of monogenic lupus, enabling clinicians to quickly determine whether genetic testing is necessary and initiate personalized treatment promptly, thereby reducing delays in diagnosis and treatment in clinical practice. Furthermore, for patients who have undergone genetic testing with negative results but score high on the nomogram, it is important not to easily dismiss the possibility of underlying issues. Further analysis, including a comprehensive review of genetic reports, whole genome sequencing or deep sequencing, and somatic cell sequencing, may be warranted to explore potential genetic mutations.

The phenotypic spectrum of monogenic lupus is broad, with different gene mutations leading to pathogenesis through distinct pathways. Currently, reports on its clinical phenotypes are scattered, with most summaries categorized based on the pathogenic pathways [25, 30,31,32,33,34,35]. Based on our statistical analysis, it has been found that the main clinical manifestations in patients with monogenic lupus were different from those observed in classical SLE.

The aforementioned literature reports that the main clinical manifestations of type I IFN diseases include skin rashes characterized primarily by chilblain-like, livedo reticularis, and nodular erythema. Neurological symptoms typically manifest as abnormal muscle tone and growth and developmental delay. Imaging findings predominantly show intracranial calcifications, white matter changes, and brain atrophy. In addition, other common features include thyroid dysfunction, hepatosplenomegaly, and pulmonary interstitial changes. Our statistical results are generally consistent, except for thyroid dysfunction, which did not show a statistically significant difference between the two groups. Classical lupus is more likely to occur in adolescent girls, whereas monogenic lupus tends to manifest in pre-adolescent children, with roughly equal gender ratios and a higher likelihood of a familial history. Regarding affected systems, monogenic lupus is more prone to chronic skin rashes, especially chilblain-like rashes, nodular erythema, and reticulated purpura. Neurological involvement is more common in monogenic lupus, with abnormal muscle tone being the most common specific manifestation, distinguishing it from classical lupus. However, other common clinical phenotypes seen in type I IFN diseases, such as subclinical thyroid dysfunction, neutropenia, and thrombocytopenia, did not show statistical significance between the two groups upon statistical analysis. Additionally, monogenic lupus more frequently presents with systemic symptoms such as fever and pulmonary interstitial changes. Some clinical features beyond lupus classification criteria are more prominent in monogenic lupus, including lymphadenopathy/hepatosplenomegaly, growth and developmental delay, elevated liver enzymes, and a history of recurrent infections. Furthermore, we observed that classical SLE is more prone to concurrent CMV infection, which is consistent with our previous team’s report [36], potentially providing additional support for the role of CMV in the pathogenesis of classical SLE [37].

A literature review assessed the application of the European Alliance of Associations for Rheumatology 2019 classification criteria in monogenic lupus, suggesting the need for the development of a simplified and effective clinical tool to enable early identification of monogenic lupus [38]. We have constructed the first predictive model for monogenic lupus using the LASSO method: predicted probability = 0.2906977 − 1.4124018 × age of onset + 15.2761768 × recurrent infections + 7.5899250 × intracranial calcifications + 3.5908622 × growth and developmental delay + 3.0491924 × abnormal muscle tone + 1.1751209 × lymphadenopathy/hepatosplenomegaly + 1.2324701 × chilblain-like skin rash. Through the model, it becomes evident that early onset is an indicator of monogenic lupus. Other clinical presentations in the model, such as a history of recurrent infections, intracranial calcifications, growth and developmental delay, abnormal muscle tone, lymphadenopathy/hepatosplenomegaly, and chilblain-like skin rash, should raise a high suspicion of monogenic lupus. Furthermore, we assessed the predictive performance of the risk score for monogenic lupus, which demonstrated good performance (AUC = 0.98, 95% CI = 0.96–1.00). Using the Youden index, we identified a cutoff score of − 9.032299, suggesting that patients with SLE who have a predictive model score greater than − 9.032299 are at a higher risk of having monogenic lupus, with a corresponding risk probability of 0.766. In other words, when the risk score is − 9.032299 or higher, there is a 76.6% probability of having monogenic lupus. We have defined a risk probability greater than 0.766 as high risk and created a nomogram chart for clinical use. This tool allows for the convenient calculation of a patient’s risk score, facilitating a rapid assessment of the probability of monogenic lupus. If the risk probability exceeds 0.766, it should raise a strong suspicion of monogenic lupus, prompting the timely completion of genetic testing and additional diagnostic evaluations including the mRNA expression of IFN-stimulated genes and ADA2 levels, among others.

The main limitations of this study include its single-center retrospective cohort design, with a majority of patients in the cohort having type I IFN diseases. Expanding the dataset and further investigating different disease types is necessary. Additionally, our patient control group did not undergo complete genetic testing. Since this study is retrospective, most of the patients who underwent comprehensive genetic testing clinically had a typical clinical presentations and received genetic testing to establish a definitive diagnosis. We are currently conducting large-scale genetic testing for lupus patients, which can be used for model validation in the next step. Another limitation is that the LASSO method may be overly conservative in selecting sparse models. Although we performed internal validation, the model has not been tested on an external dataset, so its applicability to external data should be confirmed before drawing conclusions.

In conclusion, our study found that age of onset, recurrent infections, intracranial calcifications, growth and developmental delay, abnormal muscle tone, lymphadenopathy/hepatosplenomegaly and chilblain-like skin rash are predictors of monogenic lupus. A model and a nomogram were developed and validated. This nomogram chart can assist clinical physicians in promptly identifying high-risk monogenic lupus patients during their work, thereby reducing delays in the diagnosis and treatment of patients.