基于预后梯度对IVF人群子宫内膜进行分层且不依赖子宫内膜时序
多中心前瞻性研究纳入195例接受激素替代治疗的IVF患者,对131例子宫内膜活检进行转录组测序,用AI模型分出四种与生殖预后相关的转录组模式。c1和c2预后好(最高妊娠率91%、最高持续妊娠率78%),p1和p2预后差且分别与生化流产率43%和临床流产率43%相关。
要点
- 基于AI模型可将IVF患者子宫内膜分为四种与生殖预后相关的转录组模式。
- c1模式与最高妊娠率91%相关,c2模式与最高持续妊娠率78%相关。
- p1模式与最高生化流产率43%相关,p2模式与最高临床流产率43%相关。
核心结论
- 存在四种独立于子宫内膜时序的转录组模式,与生殖预后梯度相关并涉及不同分子病理机制。
研究设计与样本
- 多中心前瞻性研究,2019年1月至2022年8月招募195例接受激素替代治疗的IVF患者,最终131例纳入分析。
- 131例中植入失败组(≥3次植入失败)32例,对照组(<3次植入失败)99例。
- 131例随机分为训练集105例和测试集26例,用于生物标志物特征发现和预测性能评估。
主要发现
- 按不良或良好生殖预后二分临床分类,激素替代治疗子宫内膜准备期间有植入失败史的患者之间无转录组区分。
- 患者被分为四种生殖预后相关模式:p1(n=24)、p2(n=14)、c2(n=32)、c1(n=61)。
- 最高妊娠率与c1相关(91%),最高持续妊娠率与c2相关(78%)。
- p1有最高生化流产率(43%),p2有最高临床流产率(43%),p1和p2均与较低妊娠率和持续妊娠率相关。
- p1与早孕期对胚胎的过度免疫反应相关,p2初始免疫耐受但后期因缺乏代谢反应而排斥胎儿。
关键数值与推荐
- 植入失败组定义为≥3次植入失败(n=32),对照组定义为<3次植入失败(n=99)。
临床怎么用
- 作者认为该转录组分类为设计更准确的子宫内膜因素不孕评估工具提供线索。
- 作者认为该分类使针对每位子宫内膜因素不孕患者制定个性化治疗策略成为可能,以提高妊娠几率。
局限性
- 由于 disrupted WOI 的异质性和不同分层组样本量有限,AI模型的人群推断有限。
- 作者指出这些有希望的发现为更大样本量的进一步临床研究提供了强有力线索。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
STUDY QUESTION: Can the disrupted window of implantation (WOI) be stratified according to transcriptomic patterns associated with reproductive success in IVF patients undergoing HRT? SUMMARY ANSWER: There are four transcriptomic patterns independent of endometrial timing associated with a gradient of reproductive prognosis underlying different molecular pathomechanisms. WHAT IS KNOWN ALREADY: A molecular heterogeneous profile independent of endometrial timing has been discovered as a cause of implantation failure that disrupt the endometrial transcriptome in the mid-secretory phase. However, the molecular heterogeneous patterns underlying the disruption remain poorly identify and understood. Characterizing the molecular heterogeneity of this endometrial disruption is crucial to develop personalized and more accurate diagnostic tools for preventive medicine, particularly for patients with a high risk of endometrial failure. STUDY DESIGN, SIZE, DURATION: In this multicenter prospective study, 195 IVF patients undergoing HRT with endometrial biopsy collection, during mid-secretory phase for endometrial progression evaluation, were recruited between January 2019 and August 2022. Out of 195 patients, 131 were finally included in the following analysis. PARTICIPANTS/MATERIALS, SETTING, METHODS: Endometrial biopsies were processed for whole endometrial transcriptome analysis using RNA-Sequencing. To identify disruptions in the WOI, the transcriptomic variation due to cyclic endometrial tissue changes was removed. Out of 195 biopsies sequenced, 131 were derived from patients that met the clinical criteria to be classified as implantation failure group (≥3 implantation failures, n = 32) or control group (<3 implantation failures, n = 99). An artificial intelligence (AI) model, based on two supervised learning algorithms: support vector machine (SVM) and k-nearest neighbors (kNN), was performed with 131 patients that were randomly allocated to training (n = 105) and test (n = 26) sets for biomarker signature discovery and assessment of predictive performance, respectively. The reproductive outcomes of the single embryo transfer immediately after biopsy collection were analyzed. Differential expression and functional analyses were performed to characterize molecular profiles. Finally, a quantitative PCR (qPCR) assay was used to corroborate the differential expression of six potential biomarkers. MAIN RESULTS AND THE ROLE OF CHANCE: With the dichotomous clinical classification of poor or good reproductive prognosis, there was no transcriptomic distinction between patients with a history of implantation failures during HRT endometrial preparation. Alternatively, using an AI model to stratify IVF patients based on the probability of endometrial disruption revealed molecular and clinical differences between patterns. Patients were stratified into four reproductive prognosis-related profiles: p1 (n = 24), p2 (n = 14), c2 (n = 32) and c1 (n = 61). The highest pregnancy rate (PR) was associated with c1 (91%) and the highest ongoing pregnancy rate (OPR) was associated with c2 (78%), linking these profiles to good reproductive prognoses. On the other hand, p1 had the highest biochemical miscarriage rate (43%) while p2 had the highest clinical miscarriage rate (43%). Notably, both p1 and p2 were related to lower PR and OPR, supporting that these profiles were associated with poor prognoses. Regarding the functional characterization in the poor prognosis profiles that were linked to miscarriages, p1 was associated with an excessive immune response against the embryo during early pregnancy stages, while p2 was initially immune-tolerant but rejected the fetus in later stages due to the lack of metabolic response. LIMITATIONS, REASONS FOR CAUTION: Due to the heterogeneous character of the disrupted WOI and the limited sample size of the different stratified groups, the AI model has limited population inference. However, our significant promising findings provide strong leads for further clinical studies with larger sample sizes. WIDER IMPLICATIONS OF THE FINDINGS: This new transcriptomic taxonomy associated with distinct reproductive outcomes provides clues to design new and more accurate evaluation tools for endometrial-factor infertility. Furthermore, it enables tailoring therapeutic strategies to apply a personalized medicine to each patient suffering from endometrial-factor infertility, improving their odds of getting pregnant. STUDY FUNDING/COMPETING INTEREST(S): This study was supported by the IVI Foundation (1706-FIVI-048-PD); Instituto de Salud Carlos III (ISCIII) and co-funded by the European Regional Development Fund "A way to make Europe" (PI19/00537 [P.D.-G.]) as well as Instituto Carlos III (ISCIII) through project (PI23/00806 [P.D.-G.]) and co-funded by European Union. Patricia Diaz-Gimeno is supported by Instituto de Salud Carlos III (ISCIII) through the Miguel Servet program (CP20/00118) co-funded by the European Union. Patricia Sebastian-Leon and Francisco Jose Sanz are funded by Instituto de Salud Carlos III (ISCIII) through the Sara Borrell postdoctoral program (CD21/00132 [P.S.-L.] and CD23/00032 [F.J.S.]) co-financed by the European Union. Josefa Maria Sanchez-Reyes was supported by a predoctoral fellowship program of the Generalitat Valenciana (ACIF/2018/072 and BEFPI/2020/028). Antonio Parraga-Leo (FPU18/01777) and Diana Marti-Garcia (FPU19/03247) were supported by predoctoral fellowship programs of the Spanish Ministry of Science, Innovation and Universities. The authors declare no conflicts of interest. TRIAL REGISTRATION NUMBER: Not applicable.
引用
- Sanchez-Reyes JM, Parraga-Leo A, Sebastian-Leon P, et al. Stratifying IVF population endometria using a prognosis gradient independent of endometrial timing†. Hum Reprod. 2025. PMID: 40796355. doi:10.1093/humrep/deaf156
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。