第一部分 通过生物信息学初探SLE与PE之间潜在的共同致病基因

目的：子痫前期（preeclampsia，PE）是一种严重的妊娠并发症，在系统性红斑狼疮（systemic lupus erythematosus，SLE）患者中更为常见。尽管这些疾病的病理生理机制尚不明确，但免疫系统在其中发挥至关重要的作用。本研究拟寻找出与SLE相关并参与PE发病的基因。

方法：从NCBI GEO数据库获取了九个微阵列数据集。使用Limma包识别差异表达基因（differentially expressed genes，DEGs），利用加权基因共表达网络分析（weighted gene co-expression network analysis，WGCNA）确定SLE数据集的关键基因，并使用Cibersort对SLE数据集进行免疫浸润分析。通过PPI网络和CytoHubba，确定前20个与PE相关的关键基因。取SLE数据集合PE数据集的关键基因的交集后构建预测PE风险的列线图，并进行受试者工作特征曲线（receiver operating characteristic，ROC）分析。

结果：PE数据集的GSEA和对DEGs富集分析表明这些基因在胎盘发育和免疫应答中起着重要作用。最终SLE数据集和PE数据集的关键基因取交集后共获得两个基因——BCL6和MME，BCL6和MME的表达水平在SLE外周血单核细胞（peripheral blood mononuclear cells，PBMCs）和PE胎盘标本中均升高。通过另外五个数据集进行外部验证，列线图也获得了良好的诊断效果（AUC：0.82-0.96）。SLE数据集的免疫浸润分析显示这两个基因的表达与中性粒细胞的凋亡和功能具有强正相关。

结论：BCL6和MME成为与SLE及PE发展相关的关键基因，我们的研究为进一步探索PE免疫相关病理生理机制提供了候选基因。

第二部分 血常规相关参数在构建系统性红斑狼疮患者子痫前期预测模型中的应用

目的：在系统性红斑狼疮（systemic lupus erythematosus，SLE）妊娠人群中筛查子痫前期（preeclampsia，PE）高风险患者。

方法：我们研究了2010年1月至2023年5月在我院分娩402例SLE患者，共计438次妊娠，单因素及多因素logistic回归分析与PE发生相关的独立危险因素，通过LASSO回归筛选不同孕周血常规及相关新型炎性指标，包括中性粒细胞淋巴细胞比值（neutrophil lymphocyte ratio，NLR）、单核细胞淋巴细胞比值（monocyte lymphocyte ratio，MLR）、血小板淋巴细胞比值（platelet lymphocyte ratio，PLR）、全身免疫炎症指数（systemic immune-inflammation index，SII）和全身炎症反应指数（systemic immune-response index，SIRI），同时考虑重复测量变量与PE之间可能存在的时变关系，拟合了三个独立PE的预测模型，并通过受试者工作特征曲线下面积（area under the receiver operating characteristic curve，AUROC）对模型进行评价。

结果：本研究SLE人群PE发生率为14.8%，其中61.5%为早发型PE。多因素分析发现孕16周前RDW水平、孕16-34周PLT计数及PLR值、孕28-34周RDW、Eos%、MCHC、MPV、P-LCR水平及SII值是PE独立预测因素。LASSO筛选变量后，孕妇年龄、规律产检、慢性高血压、APS、孕期使用羟氯喹、降压药和免疫抑制剂/免疫球蛋白、早孕期尿沉渣、24h尿总蛋白水平、抗ds-DNA抗体及孕16周前的HGB水平等变量构成了任何孕周发生的PE风险预测模型，该模型的AUC值为0.90，95% CI: 0.86-0.94。预测28周后的PE的模型预测变量包括规律产检、慢性高血压、APS、孕期使用羟氯喹、降压药和免疫抑制剂/免疫球蛋白、早孕期尿沉渣和24h尿总蛋白水平，AUC值为0.88, 95% CI: 0.82-0.93，表明该模型可由模型1替代。最终纳入到晚发型PE风险预测模型的变量有规律产检、慢性高血压、APS、降压药和免疫抑制剂/免疫球蛋白的使用、早孕期WBC和SIRI水平及孕28-34周测定的MCHC水平，模型的AUC=0.91, 95% CI: 0.86-0.97。

结论：SLE人群EOPE发生率更高，除SLE相关特异性指标外，血常规参数及其相关炎性指标也可用于预测PE，且模型中加入血常规相关参数可提高模型的预测效能，帮助临床医生通过早期干预改善高危女性的妊娠结局。

第三部分 常规产检指标在一般人群早发型和晚发型子痫前期中的差异及在构建子痫前期及其亚型预测模型的应用

目的：比较不同亚型PE与对照组间妊娠各阶段常规实验室各指标间的差异，并构建PE及其亚型的预测模型。

方法：纳入2018年1月至2023年12月在我院分娩的499名单胎非免疫病孕妇，分为三组：对照组275例，PE组224例，包括早发型PE（early-onset preeclampsia，EOPE）96例，晚发型PE（late-onset preeclampsia，LOPE）128例。比较各组的外周血参数，包括妊娠各阶段的血常规、凝血、肝肾功、早孕期子宫动脉频谱、中孕期铁代谢和血脂等，以发现EOPE和LOPE之间的差异。利用多重插补后的妊娠早期（<16周）临床资料，将数据分为训练集和验证集，通过基于stepwise的多因素logistic回归分析构建PE、EOPE、LOPE预测模型，并绘制列线图可视化模型，绘制校准曲线和受试者工作特征曲线（receiver operating characteristic，ROC），计算曲线下面积（area under curve，AUC）和Hosmer-Lemeshow检验对模型进行评价。

结果：（1）血常规方面，与对照组相比，妊娠早期，LOPE组PLT计数升高及相关血小板参数，如MPV、PDW和P-LCR降低；孕16-34周，EOPE组异常更为显著，表现为HGB、HCT升高，PLT计数及相关炎性指标降低，伴血小板相关参数降低为主，而LOPE组则主要表现为PLT计数及LY%升高为主。（2）血生化及凝血方面，与对照相比，妊娠早期，各指标在EOPE组无明显差异，而在孕16-34周时两组间均存在显著差异（p均 < 0.001），而LOPE组与对照组两组在妊娠早期及孕16-34周的差异类似，均为UA、cCa及Fbg水平在LOPE组升高。（3）妊娠早期平均子宫动脉搏动指数及子宫动脉舒张早期切迹比例仅在EOPE组明显更高（p均=0.002）。妊娠中期血脂水平在对照组与LOPE和EOPE组间差异不明显，但妊娠中期铁代谢各指标在EOPE组存在显著异常，Fe、TS和Fer在EOPE组明显升高（p=0.009，p<.001和p<.001），相应的TRF和TIBC在显著降低（p=0.005和0.001）。（4）终止妊娠前，EOPE组与LOPE相比，生化和凝血指标均存在显著差异，血常规指标中仅WBC计数在EOPE组明显升高，而血小板相关参数和RDW-C在EOPE组显著降低。（5）最终的PE风险预测模型中共包括13个变量，在训练集和测试集中的AUC值分别为0.73，95% CI（0.71-0.75）和0.71，95% CI（0.68-0.75）。EOPE风险预测模型包含8个变量，在训练集的AUC为0.78，95% CI（0.75-0.81），敏感度为80.7%，特异性为67.4%，PPV为46.3%，NPV为90.9%，测试集的AUC为0.81，95% CI（0.76-0.85），但校准度较差。LOPE预测模型也由8个变量构成，该模型在训练集和测试集中的AUC值分别为0.75，95%CI（0.72-0.77）和0.74，95% CI（0.70-0.78）。

结论：在这项研究中，我们比较了对照组与不同孕周发病PE组间的外周血相关参数，发现EOPE和LOPE可能存在不同的发病机制，依据母体一般特征及妊娠16周前化验检查构建的EOPE风险预测模型区分度最好，但校准度较差，需扩大样本进一步验证。

Part I: Identification of lupus-associated genes in the pathogenesis of preeclampsia via bioinformatic analysis

Objective: Preeclampsia (PE), a potentially fatal pregnancy complication, is more common in patients with systemic lupus erythematosus (SLE). Although the pathophysiology of these illnesses is uncertain, the immune system is essential. We looked for genes linked to SLE involved in the etiology of PE.

Methods: We obtained nine microarray datasets from the NCBI GEO database. Limma was used to identify the DEG. We employed weighted gene coexpression network analysis (WGCNA) to identify the hub genes of the SLE and examined immune infiltration using Cibersort. Using the PPI network and CytoHubba, the top 20 PE hub genes were found. Following the construction of a nomogram and receiver operating characteristic (ROC) analysis to predict the risk of PE.

Results: GSEA and PE DEGs enrichment analysis indicated a major role in placenta development and immune response. We acquired two pivotal genes, BCL6 and MME, and confirmed their validity using five datasets. Good diagnostics were obtained from the nomogram (AUC: 0.82-0.96). The expression levels of both genes were elevated in SLE peripheral blood mononuclear cells (PBMCs) and PE placental specimens within the case group. Immune infiltration analysis of the SLE dataset revealed a strong positive correlation between the expression of both genes and neutrophil infiltration.

Conclusion: BCL6 and MME emerged as key genes in lupus-related pregnancies associated with the development of PE, for which we devised a nomogram. These data provide candidate genes for further exploration in the diagnosis and understanding of the pathophysiology of PE.

Part II: The application of hematological parameters in constructing a predictive preeclampsia (PE) predictive model in patients with systemic lupus erythematosus (SLE).

Objective: To screen for a high risk of preeclampsia (PE) in women with systemic lupus erythematosus (SLE).

Methods: We conducted a study on 402 SLE patients who gave birth at our hospital from January 2010 to May 2023, with a total of 438 pregnancies. We analyzed single-factor and multiple-factor logistic regression to identify independent risk factors associated with PE. Through LASSO regression, we screened complete blood count (CBC) parameters and novel inflammatory markers, including neutrophil-lymphocyte ratio (NLR), monocyte-lymphocyte ratio (MLR), platelet-lymphocyte ratio (PLR), systemic immune-inflammation index (SII), and systemic immune-response index (SIRI), at different trimesters. Considering the possible time-varying relationship between repeated measurement variables and PE, three independent predictive models for PE were fitted and evaluated using the Area Under the Receiver Operating Characteristic Curve (AUROC).

Results: The PE incidence was 14.8% in SLE pregnant patients, with 61.5% being early-onset PE (EOPE). Multifactorial analysis revealed that RDW before 16 weeks of gestation, PLT and PLR between 16-34 weeks of gestation, RDW, Eos%, MCHC, MPV, P-LCR, and SII during 28-34 weeks of gestation are independent predictive factors for PE. After LASSO regression, the predictive model for PE diagnosed at any gestational week (Model 1) included maternal age, regular prenatal care, chronic hypertension (HTN), APS, usage of hydroxychloroquine (HCQ) during pregnancy, antihypertensive drugs and immunosuppressive drugs, positive urine sediment in early pregnancy, 24-hour urine protein (24hUP), anti-dsDNA antibodies, and HGB before 16 weeks of gestation. The AUC value for this model was 0.90 with a 95% confidence interval (CI) of 0.86-0.94. For the prediction of PE diagnosed after 28 weeks of gestation, model 2 included regular prenatal care, HTN, APS, usage of HCQ during pregnancy, antihypertensive drugs and immunosuppressive drugs, urine sediment in early pregnancy, and 24hUP. The AUC value for this model was 0.88, 95% CI 0.82-0.93, indicating that this model can be replaced by Model 1. Finally, variables included in the late-onset PE predictive model were regular prenatal care, HTN, APS, usage of antihypertensive drugs and immunosuppressive drugs, WBC and SIRI in the first trimester, and MCHC measured between 28-34 weeks of gestation. The model's AUC was 0.91 with a 95% CI of 0.86-0.97.

Conclusion: The incidence of EOPE is higher in the SLE population. Besides SLE-specific markers, hematological parameters and related inflammatory markers can also be used to predict preeclampsia (PE). Moreover, incorporating hematological parameters into the model improves its predictive performance, assisting clinicians in early intervention to improve pregnancy outcomes in high-risk women.

Part III: Exploring the factors related to early-onset and late-onset preeclampsia based on routine prenatal examination indicators and constructing a predictive model for preeclampsia and its subtypes.

Objective: To compare the differences in routine laboratory parameters during different stages of pregnancy between different subtypes of preeclampsia (PE) and the control group, and to construct predictive models for PE and its subtypes.

Methods: A total of 499 singleton pregnancies without immune diseases, who were delivered in our hospital from January 2018 to December 2023, were included and divided into three groups: control group (275 cases), PE group (224 cases), including early-onset PE (EOPE, 96 cases) and late-onset PE (LOPE, 128 cases). Peripheral blood parameters were compared among the groups, including complete blood count (CBC), coagulation function, liver and kidney function, uterine artery Doppler spectrum in early pregnancy, iron metabolism, and lipid profiles in the second trimester, to discover the differences between EOPE and LOPE. Clinical data from early pregnancy (<16 weeks) after multiple imputations were divided into training and validation sets. Multivariable logistic regression analysis based on stepwise selection was performed to construct predictive models for PE, EOPE, and LOPE. The models were visualized using nomograms. The performance of the model was evaluated for the receiver operating characteristic (ROC) curve, calibration, and the Hosmer-Lemeshow test.

Results: (1) In terms of CBC, compared to the control group, during early pregnancy, the LOPE group exhibited increased PLT count and decreased related platelet parameters such as MPV, PDW, and P-LCR. During 16-34 weeks of gestation, the EOPE group showed more significant abnormalities, characterized by increased HGB and HCT, decreased PLT count, and related inflammatory markers, primarily accompanied by a decrease in platelet-related parameters. In contrast, the LOPE group mainly demonstrated increased PLT count and LY% as the primary findings. (2) In terms of blood biochemistry and coagulation, compared to the control group, there were no significant differences in various indicators during early pregnancy in the EOPE group. However, significant differences were observed between the two groups during 16-34 weeks of gestation (p < 0.001). The differences between the LOPE group and the control group were similar in early pregnancy and 16-34 weeks of gestation, with UA, cCa, and Fbg levels increasing in the LOPE group. (3) During early pregnancy, the mean pulsatility index and early diastolic notch ratio of the uterine artery were significantly higher only in the EOPE group (p = 0.002). In the second trimester, there were no significant differences in lipid levels between the control group and the LOPE and EOPE groups. However, significant abnormalities in iron metabolism parameters were evident in the EOPE group, with significantly higher levels of serum Fe, TS, and Fer (p = 0.009, p < 0.001, and p < 0.001, respectively), and corresponding significantly lower levels of TRF and TIBC (p = 0.005 and p = 0.001, respectively). (4) Before termination of pregnancy, significant differences in biochemical and coagulation indicators were observed between the EOPE and LOPE groups. Among the CBC, white blood cell (WBC) count was significantly higher in the EOPE group, while platelet-related parameters and RDW-C were significantly lower. (5) The final PE risk prediction model included a total of 13 variables, with AUC values of 0.73 (95% CI: 0.71-0.75) in the training set and 0.71 (95% CI: 0.68-0.75) in the testing set. The EOPE risk prediction model consisted of eight variables, with an AUC of 0.78 (95% CI: 0.75-0.81) in the training set, sensitivity of 80.7%, specificity of 67.4%, positive predictive value (PPV) of 46.3%, negative predictive value (NPV) of 90.9%, and an AUC of 0.81 (95% CI: 0.76-0.85) in the testing set, but with poor calibration. The LOPE prediction model also comprised eight variables, with AUC values of 0.75 (95% CI: 0.72-0.77) in the training set and 0.74 (95% CI: 0.70-0.78) in the testing set.

Conclusion: In this study, we compared peripheral blood parameters during different stages of pregnancy between the control group and PE groups of different subtypes. We found that EOPE and LOPE may involve different pathogenic mechanisms. The EOPE risk prediction model, based on maternal general characteristics and laboratory tests before 16 weeks of gestation, had the best discrimination but poor calibration. Further validation with larger sample sizes is necessary.