【作者】 李超；

【导师】 于志勇；

【作者基本信息】 山东大学 ， 肿瘤学（专业学位）， 2025， 博士

【摘要】 背景新辅助化疗(Neoadjuvant chemotherapy,NAC)是乳腺癌的重要治疗方案之一,其价值在于:缩小肿瘤体积、降低淋巴结分期,提高保乳及保腋窝概率;体内评估肿瘤组织对治疗的敏感度,指导辅助治疗方案选择,提高治疗预后。然而,乳腺癌异质性高的特点使得患者对NAC的治疗反应存在显著差异,约30%-60%的患者无法取得病理完全缓解(Pathological complete response,pCR),因此亟需开发精准的生物标志物或预测模型以筛选NAC敏感人群,从而优化治疗决策。目前研究已从临床特征、功能影像学特征及分子生物学指标等维度探索了 NAC治疗反应的预测因素。例如,基因组学技术的进步推动了多基因评分模型(如Oncotype DX、MammaPrint)的开发,但这些基因评分模型多基于小样本量、单一组学平台的数据构建,且未整合临床病理特征(如肿瘤分级、激素受体状态)及动态治疗反应信息,导致预测效能受限。这一局限性凸显了构建多维度、跨组学预测模型的必要性,需纳入大规模临床队列的纵向数据,并结合组织基因组测序与动态生物标志物监测,以提高模型的临床适用性。在此背景下,循环肿瘤DNA(Circulatingtumor DNA,ctDNA)具备的非侵入性和反映即刻肿瘤分子生物信息的优势使其受到广泛关注。ctDNA是肿瘤细胞进入到循环系统的DNA片段,其丰度与肿瘤负荷呈正相关,多时间节点检测可动态追踪治疗过程中特异性突变(如TP53、PIK3CA)的清除率,为早期评估NAC疗效提供分子依据。然而,ctDNA检测易受克隆性造血或正常细胞游离DNA干扰,且仅反映血液中肿瘤异质性的部分信息。因此,联合组织DNA测序(揭示肿瘤基线突变谱)与ctDNA动态监测,可以更全面地评估NAC反应。值得注意的是,目前尚缺乏整合组织-液体活检的多组学研究,这一空白领域亟需系统性探索。进一步研究发现,肿瘤免疫微环境作为动态调控NAC反应的关键枢纽,其作用机制逐渐被揭示。三阴性乳腺癌(Triple-negative breast cancer,TNBC)因其高异质性和高免疫原性的特征,使其成为研究TIME与化疗协同机制的理想对象。肿瘤浸润淋巴细胞(Tumor infiltrating lymphocyte,TILs)的高密度浸润与TNBC患者较高的pCR率显著相关,这可能归因于细胞毒性T细胞(CD8+T细胞)介导的免疫杀伤作用与化疗的协同效应。然而,TILs的异质性构成(包含Treg、TAM等细胞)可能产生免疫抑制效应,提示需要解析特定免疫细胞亚群(如PD-1+CD8+T细胞)的分布及功能状态,明确影响NAC治疗反应的特定免疫指标及免疫细胞亚群。本研究基于上述多维度预测模型构建及机制探索的科学需求,设计了三阶段递进式研究方案:第一部分(多组学预测模型构建):纳入接受NAC的乳腺癌患者队列,采集治疗前组织进行靶向基因组测序(涵盖乳腺癌驱动基因及治疗相关通路基因panel),整合临床病理特征与突变谱数据,通过机器学习算法构建NAC治疗反应及无病生存期的预测模型;第二部分(ctDNA动态检测验证与模型优化):在NAC治疗周期内(基线、2周期、4周期及术后)开展多时间节点ctDNA动态监测来预测NAC治疗反应,并将其与第一部分模型整合,构建更高效能的治疗反应及预后预测模型;第三阶段(特定免疫细胞亚群功能研究):基于第一部分组织测序信息及基因表达综合数据库(Gene Expression Omnibus,GEO)数据库TNBC队列的转录组数据,筛选与pCR显著相关的免疫特征基因及免疫细胞亚群,利用临床样本验证免疫指标及免疫细胞亚群与NAC治疗反应关系,进一步通过动物试验验证免疫指标及免疫细胞亚群对TNBC化疗效果的影响及其机制。目的1.通过整合组织DNA突变测序及临床病理特征构建模型来预测乳腺癌NAC治疗反应及预后。2.通过NAC期间ctDNA突变动态变化,及其整合构建新模型来预测乳腺癌NAC治疗反应及预后。3.筛选影响TNBC患者NAC效果的靶点ICOS以及ICOS+T细胞亚群,通过ICOS激动剂增强TNBC化疗效果,并探究其机制。方法第一部分:基于组织DNA突变测序及临床病理特征构建乳腺癌NAC治疗反应及预后预测模型纳入山东省肿瘤医院及聊城市人民医院乳腺外科于2016年1月至2020年4月诊治的接受NAC的女性乳腺癌患者243例,收集临床病理信息,利用患者NAC前活检组织进行靶向基因组测序,筛选与治疗反应相关的DNA突变特征及临床病理因素,构建NAC治疗反应及预后预测模型,并验证模型的效能。第二部分:循环肿瘤DNA及其构建模型预测乳腺癌NAC治疗反应及预后纳入山东省肿瘤医院第一部分同期的接受NAC的患者56例,其中53例为第一部分重叠病例。在NAC过程中动态采集多时间点的血样进行ctDNA突变检测,动态分析各个时间点ctDNA状态及变化与NAC治疗反应的相关性。将ctDNA状态作为独立因素加入到第一部分预测模型,构建新的预测模型,并验证模型的预测性能及预后风险分层。第三部分:靶向ICOS调控CD8+T细胞功能增强TNBC化疗效果的研究通过整合前期组织突变谱(第一部分)与GEO数据库TNBC队列的转录组数据来筛选TNBC特异性免疫调控靶点,发现可诱导共刺激因子(Inducible T cell co-stimulator,ICOS)在化疗敏感组显著高表达,利用临床标本及数据库信息验证了 ICOS表达与NAC后pCR的关系,通过多重免疫荧光技术定量分析ICOS+T细胞富集程度与pCR的相关性。建立小鼠的移植瘤模型,通过ICOS激动剂验证其对TNBC化疗效果的影响,并探究其作用机制。结果第一部分:基于组织DNA突变测序及临床病理特征构建乳腺癌NAC治疗反应及预后预测模型1.组织DNA突变及临床病理特征与NAC治疗反应显著相关本部分纳入研究共243例患者,共总结了 192份高频突变≥10%的体细胞突变样本,鉴定出425个独特基因,高突变基因前5位分别为TP53、KMT2C、PIK3CA、EPHA1和EPPK1。按照NAC后pCR和非pCR进行差异突变基因分析,筛选出与pCR状态显著相关的5个单核苷酸变异(Singlenucleotide variant,SNV)和4个拷贝数变异(Copy numbervariation,CNV)。通过单因素分析与pCR相关的临床病理特征,筛选得到了 3个临床病理因素(Luminal A、Her2+和Ki-67)。2.机器学习整合组织DNA突变信息和临床病理因素构建的NAC治疗反应预测模型效能良好运用机器学习,将9个突变基因(5个SNV和4个CNV)和3个临床因素(Luminal A、Her2+和Ki67)进行整合,得到的组合模型在训练集(Area Under Curve,AUC:0.871,95%Confidence Interval,CI:0.797-0.927)、验证集(AUC:0.771,95%CI:0.649-0.883)和外部验证集(AUC:0.726,95%CI:0.556-0.865)具有很高的敏感性和特异性,优于其他组合模型。3.NAC治疗反应模型可以很好地预测预后生存随访分析发现,NAC治疗反应模型对患者的DFS结局有较好的预测效果,且长期预后的预测效果优于短期预测效果(1年时AUC:0.749;3年时AUC:0.830)。计算中位风险评分,并依此分成高危组和低危组,低危组的DFS显著高于高危组(P<0.0001)。第二部分:循环肿瘤DNA及其构建模型预测乳腺癌NAC治疗反应及预后1.ctDNA的动态清除能够预测NAC的治疗反应本部分纳入共56例患者,T0时46%的患者ctDNA阳性,TNBC患者ctDNA阳性比例(80%)高于其他亚型,NAC期间ctDNA阳性率随着治疗进行而逐渐降低,从T0时的46%下降到T1时的14%,T2时为13%,T3时为10%。所有pCR患者在T2和T3时均达到ctDNA阴性,T2和T3时ctDNA 阳性的患者都没有达到pCR。T1时ctDNA阳性的患者中,高达85.8%(6/7)的患者未达到pCR,而T1时ctDNA转阴性的患者中有69%(9/13)未达到pCR。T2和T3时,ctDNA 阳性的患者均未达到pCR(4/4),而ctDNA转阴性的患者中也有69%(11/16)未达pCR。2.ctDNA的动态变化与预后显著相关本部分分析了 ctDNA动态变化和DFS的相关性,T3时ctDNA仍阳性的患者的复发风险明显高于T0、T1、T2和T3时ctDNA阴性的患者,与T0时ctDNA阴性的患者(N=22)相比,T1、T2或T3时ctDNA转阴性的患者(N=18)与之有相似的复发风险。T1、T2、T3时ctDNA转阴性者比T3 ctDNA仍阳性者的DFS更长。T3时ctDNA转阴性比率与生存率的提高有关。根据NAC后pCR情况和ctDNA状态对患者进行分层,发现7例达到pCR且ctDNA阴性患者具有显著延长的DFS,在未达到pCR的38例患者中,ctDNA 阳性与较差的DFS相关。未能达到pCR但ctDNA阴性的患者复发风险与达到pCR的患者相似。3.整合ctDNA动态监测与组织DNA突变及临床病理特征的多因素预测模型具有更好的治疗反应及预后预测效能将第一部分构建的预测模型与动态ctDNA状态相结合构建新模型,结果显示,结合后的模型可以更好的预测NAC治疗反应。加入T0时ctDNA状态与加入T0和T1时、加入T0、T1和T2时ctDNA状态的模型,具有相同的高灵敏度和特异性,AUC值都达到0.961。新模型对NAC患者的DFS结局也有较好的预测效能,且长期预后的预测效果优于短期预测效果(1年时AUC=1.000,2年时AUC=0.941)。计算中位风险评分,并依此分成高危组和低危组,低危组的DFS显著高于高危组(P=0.0031)。第三部分:靶向ICOS调控CD8+T细胞功能增强TNBC化疗效果的研究1.TNBC患者NAC治疗反应与免疫细胞浸润程度呈正相关本部分收集了第一部分的接受NAC的TNBC患者队列42例,分析了患者的临床因素和组织DNA突变情况,发现达到pCR患者的间质肿瘤浸润淋巴细胞高于非pCR患者。对GSE25055数据集的96例接受NAC的TNBC患者分析,发现pCR与非pCR组样本免疫细胞丰度无差别,但非pCR样本中CD8+na(?)ve T细胞(静息)、Treg显著高于pCR样本。2.ICOS及ICOS+T细胞与TNBC患者NAC治疗反应及预后相关通过GSE25055数据集中TNBC患者pCR与非pCR的转录组差异基因与DNA测序CNV突变基因取交集。交集基因的通路富集到免疫相关的“T细胞受体信号通路”,且其中关键基因为ICOS,ICOS表达与免疫评分呈正相关(P<0.001)。临床组织免疫组化结果显示pCR患者的ICOS表达显著高于非pCR患者,多重免疫荧光染色分析发现pCR患者穿刺肿瘤组织中CD8+T细胞、CD4+T细胞和ICOS+T细胞明显高于非pCR组。TCGA数据库分析了发现TNBC中ICOS高表达的患者具有更好的OS(P=0.012)。3.ICOS激动剂增加TNBC化疗效果为了验证ICOS及其免疫通路对TNBC化疗效果的影响,构建了 4T1和E0771细胞株荷瘤小鼠模型,结果发现与紫杉醇组或ICOS激动剂组单因素治疗相比,紫杉醇+ICOS激动剂联合治疗组的肿瘤明显较小,证实了 ICOS激动剂能够增强紫杉醇对TNBC细胞株的杀伤作用。4.ICOS激动剂通过增强CD8+T细胞功能增加TNBC化疗效果为了探究ICOS激动剂影响紫杉醇效果具体发挥作用的免疫细胞亚群及机制,我们对接受不同处理(Control组;紫杉醇组;ICOS激动剂组;紫杉醇+ICOS激动剂联合组)后肿瘤组织进行流式细胞分析免疫细胞分布的差别,发现与紫杉醇组或ICOS激动剂组单因素治疗相比,联合给药时,CD8+T淋巴细胞比例以及ICOS+CD8+T/CD8+T淋巴细胞比例明显增加,并且CD8+T淋巴细胞的效应分子细胞亚群(TNFα+和Granzyme B+T淋巴细胞)显著富集。结论1.基于组织DNA突变测序及临床病理特征构建的乳腺癌NAC治疗反应预测模型,具有很高的敏感性和特异性,并且该模型也可以很好地预测预后。2.ctDNA治疗期间的变化能够反映NAC的治疗效果,将其与组织DNA突变测序及临床病理特征整合构建的NAC治疗反应及预后预测模型,具有更高的敏感度和特异性。3.ICOS及ICOS+T细胞亚群水平与TNBC患者NAC治疗反应及预后密切相关。ICOS激动剂通过增强CD8+T细胞效应因子TNF-α和Granzyme B产生,发挥其抗肿瘤效应,提高TNBC化疗效果。更多还原

【Abstract】 BackgroundNeoadjuvant chemotherapy(NAC)is the one of the preferred treatments for breast cancer.Its value lies in:reducing tumor and lymph node staging to improve the probability of breast and axillary lymph node conservation surgery;evaluating the sensitivity of chemotherapy in vivo to guide adjuvant treatment strategies and improve prognosis.However,the high heterogeneity of breast cancer results in significant differences in the treatment response to NAC among patients,about 30%-60%of patients fail to achieve pathological complete response(pCR).It is necessary to develop accurate biomarkers or predictive models to screen NAC sensitive population and guide the precise treatment.At present,the predictive factors of NAC treatment response have been explored from the dimensions of clinical features,functional imaging features and molecular biological indicators.For example,advances in genomics technology have led to the development of multigene score models(e.g.,Oncotype DX,MammaPrint),but most of these gene score models are based on small sample sizes,single omics platform data,and do not integrate clinicopathological characteristics(e.g.,tumor grade,hormone receptor status)and dynamic treatment response information.Therefore,the prediction efficiency is limited.This limitation highlights the need to construct multi-dimensional,cross-omics prediction models that incorporate longitudinal data from large clinical cohorts and integrate tissue genome sequencing and dynamic biomarker monitoring to improve the clinical applicability of models.In this context,circulating tumor DNA(ctDNA)has received widespread attention because of non-invasive and real-time reflection of tumor molecular characteristics.ctDNA is derived from free DNA fragments of tumor cells,and its abundance is positively correlated with tumor burden.Multi-time node detection can dynamically track the clearance rate of specific mutations(such as TP53 and PIK3CA)during treatment,which can provide molecular basis for early evaluation of NAC efficacy.However,ctDNA detection is susceptible to interference by clonal hematopoietic or normal cell-free DNA,and only reflects partial information about tumor heterogeneity in blood.Therefore,the combination of tissue DNA sequencing,which reveals the baseline mutation spectrum of the tumor,with ctDNA dynamic monitoring may provide a more comprehensive assessment of NAC response.There is still a lack of multi-omics studies integrating tissue-liquid biopsy,and this blank field needs systematic exploration.Further studies have found that the tumor immune microenvironment is a key hub for dynamic regulation of NAC response,and its mechanism has been gradually revealed.For example,triple-negative breast cancer with high-density infiltration of tumor infiltrating lymphocytes was related to higher pCR.This may be attributed to the synergistic effect of cytotoxic T cells-mediated immune killing and chemotherapy.However,the heterogeneous composition of TILs,such as regulatory T cells and tumor macrophages,may produce immunosuppressive effects,suggesting that it is necessary to analyze the distribution and functional status of specific immune cell subsets,such as PD-1+CD8+T cells,and to clarify the specific immune parameters and immune cell subsets that affect the response to NAC treatment.Based on the above-mentioned scientific needs of multi-dimensional prediction model construction and mechanism exploration,this study designed a three-stage progressive research scheme:Part Ⅰ(Multi-omics prediction model construction):the patients with breast cancer received NAC was enrolled,and tumor tissue samples before treatment were collected for targeted genomic sequencing(including breast cancer driver gene and therapy-related pathway gene panel).The clinicopathological features and mutation spectrum data were integrated,and the NAC treatment response and disease-free survival prediction models were constructed by machine learning algorithms.Part Ⅱ(Validation and model optimization):Dynamic ctDNA monitoring was performed at multiple time points during the NAC treatment cycle(baseline,cycle 2,cycle 4 and after NAC)to predict NAC treatment response.This dynamic monitoring was integrated with the first part of the model to construct a more efficient prediction model for treatment response and prognosis.Phase Ⅲ(Functional study of specific immune cell subsets):Based on the first part of the tissue sequencing dataset and the transcriptome data of the TNBC cohort from the Gene Expression Omnibus(GEO)database,the immune signature genes and immune cell subsets significantly related to pCR were screened.The relationship between immune indexes and immune cell subsets and NAC treatment response was verified by clinical samples,and the effect and mechanism of immune indexes and immune cell subsets on the chemotherapy effect of TNBC were further verified by animal experiments.Objectives1.To construct predictive model of response to NAC and prognosis for breast cancer by integrating DNA mutation sequencing with clinicopathological features.2.To predict NAC response and prognosis by dynamic detection of ctDNA and construct predictive model with tissue DNA and ctDNA.3.To find the target ICOS and ICOS+T cell subsets that affect the efficacy of NAC in TNBC patients,and to enhance the chemotherapy effect of TNBC by ICOS agonist,and to clarify its mechanism.MethodsPart Ⅰ:Construction of a model based on DNA mutation sequencing and clinicopathological features to predict response to NAC and survival in breast cancer243 female breast cancer patients who received NAC in Shandong Cancer Hospital and Liaocheng People’s Hospital from January 2016 to April 2020 were enrolled.Clinicopathological information was collected.Targeted genomic sequencing was performed on biopsy tissues before NAC to screen DNA mutation characteristics and clinicopathological factors related to treatment response.Prediction model of NAC response and prognosis was constructed and verified.Part Ⅱ:Construction of a model with tissue DNA mutation sequencing and dynamic ctDNA to predict response to NAC and survival in breast cancerFifty-six female breast cancer patients received NAC in Shandong Cancer Hospital in same time were enrolled,of which 53 cases were part I overlapping cases.Blood samples were collected dynamically at multiple time points during NAC for ctDNA mutation detection,and the correlation between ctDNA status and changes at each time point and NAC treatment response was analyzed dynamically.The ctDNA status was added into the first part of the prediction model as an independent factor to construct a new prediction model,and the predictive performance and prognostic risk stratification of the model were verified.Part Ⅲ:Targeting ICOS to regulate CD8+T cell function and enhance the chemotherapy effect of TNBCBy integrating the previous tissue mutation profile(Part Ⅰ)and the transcriptome data of the TNBC cohort in the GEO database to screen the TNBC specific immune regulatory targets,it was found that the inducible T cell co-stimulator(ICOS)was significantly highly expressed in the chemotherapy-sensitive group.The relationship between ICOS expression and pCR after NAC was verified by clinical specimens and database information.The correlation between the enrichment degree and spatial distribution of ICOS+T cells and pCR was quantitatively analyzed by multiple immunofluorescence techniques.A mouse xenograft tumor model was established to verify the effect of ICOS agonist on the chemotherapy effect of TNBC,and to explore its mechanism.ResultsPart Ⅰ:Construction of a model based on DNA mutation sequencing and clinicopathological features to predict response to NAC and DFS in breast cancer1.Tissue DNA mutation and clinicopathological features were significantly correlated with NAC treatment responseA total of 243 patients were enrolled.192 somatic mutation samples with high frequency mutations ≥10%were analyzed and summarized.A total of 425 genes were found,and the top five frequency mutated genes were TP53,KMT2C,PIK3CA,EPHA1 and EPPK1.High frequency mutations in 192 patients were analyzed according to pathological complete response(pCR)and non-pCR.Five single nucleotide variant(SNV)mutations and four copy number variation(CNV)mutations that related to pCR were screened out.Univariate analysis showed that luminal A,Her2+and Ki67 were associated with pCR.2.A predictive model incorporating DNA mutation status and clinicopathological features can predict response to NAC treatment with good performanceWe constructed a predictive NAC response mode including nine mutated genes and three clinical factors.The model achieved good performance in the training set(AUC:0.871,95%CI:0.797-0.927),validation set(AUC:0.871,95%CI:0.797-0.927),0.771,95%CI:0.649-0.883)and the external validation set(AUC:0.726,95%CI:0.556-0.865)had high sensitivity and specificity,which were better than other combination models.3.Response predictive models can predict survival wellThe model had a good performance on the DFS of patients,and the prediction performance of 3 years DFS was better than that of 1 year DFS(AUC=0.749 at 1 year and AUC=0.830 at 3 years).Low-risk group showed higher DFS than the high-risk group(P<0.0001).Part Ⅱ:Construction of a model with tissue DNA mutation sequencing and dynamic ctDNA to predict response to NAC and DFS in breast cancer1.Dynamic clearance of ctDNA can predict response of NACA total of 56 patients enrolled.At baseline(T0),46%of patients were ctDNA positive,and the ctDNA decreased gradually with the duration of NAC.It decreased from 46%at T0 to 14%at T1,13%at T2,and 10%at T3.All pCR patients had undetectable ctDNA at T2 and T3,and none with positive ctDNA at T2 and T3 achieved pCR.Among the patients with positive ctDNA at T1,85.8%(6/7)did not achieve pCR,while 69%(9/13)of the patients with negative ctDNA at T1 did not achieve pCR.At T2 and T3,none of the patients with positive ctDNA achieved pCR(4/4),and 69%of the patients with negative ctDNA also failed to achieve PCR(11/16).2.The dynamic status of ctDNA has a significant relationship with prognosisWe analyzed the relationship between dynamic ctDNA status and DFS,and found that patients with positive ctDNA at T3(n=5)showed significantly higher recurrence risk than that with negative ctDNA at T0,T1,T2 and T3(p=0.027).Patients with negative ctDNA at T1,T2,or T3(n=18)had a similar recurrence risk as those with negative ctDNA at T0(n=22).Patients with negative ctDNA had longer DFS than those with positive ctDNA at T3.Seven cases(100%)showed good DFS by pCR(all ctDNA negative).Among patients who did not achieve pCR(n=38),ctDNA positivity(n=5)was associated with worse DFS.The risk of metastasis in patients who failed to achieve pCR but were ctDNA-negative was similar to that in patients who achieved pCR.3.The multivariate prediction model integrating ctDNA dynamic monitoring with tissue DNA mutation and clinicopathological features has better predictive performanceWe combined the dynamic ctDNA status with the prediction model in first part to construct a new model.The combined model could better predict the pCR status after NAC.The ctDNA status at T0 had the same high sensitivity and specificity as that at T0 and T1,and ctDNA status at T0,T1 and T2.The AUC values all reached 0.961,indicating that a good prediction level could be obtained as long as the ctDNA state at T0 was added.This model showed a better prediction performance on DFS of patients.The 2-years DFS prediction was better than the one-year DFS prediction(AUC=1.000 at 1 year and AUC=0.941 at 2 years).Part Ⅲ:Targeting ICOS to regulate CD8+T cell function and enhance the chemotherapy effect of TNBC1.Response to NAC in TNBC is correlated with the immune cell infiltrationA total of 42 TNBC patients in the part I of cohort were enrolled.The clinical factors and tissue DNA mutations of the patients were analyzed.It was found that the TILs in the patients who achieved pCR were better than those in non-pCR patients.Analysis of 96 TNBC patients who received NAC in GSE25055 dataset showed that there was no difference in the abundance of immune cells between the pCR and non-pCR group,and the specific types and content of immune cells may be different.CD8+naive T cells(resting)and Treg in non-pCR samples were higher than pCR group.2.ICOS and ICOS+T cells are related to NAC response and prognosis in TNBC patientsThe intersection of transcriptome differential genes in GSE25055 dataset and CNV mutant in DNA mutation sequencing in TNBC was taken.The pathways of intersection genes were enriched in immune-related "T cell receptor signaling pathway",and the key gene was ICOS.The ICOS level was correlated with the immune score(P<0.001).Clinical samples showed that the expression of ICOS in pCR patients was significantly higher than that in non-pCR patients.Multi-color immunofluorescence staining analysis showed that CD8+T cells,CD4+T cells and ICOS+T cells in pCR patients were significantly higher than those in non-pCR group.TCGA database found that TNBC with high level ICOS had better OS than that of low level ICOS(P=0.012).3.ICOS agonists increase the efficacy of chemotherapy in TNBCTo verify the effect of ICOS and ICOS+T cell subsets on the efficacy of chemotherapy in TNBC,4T1 and E0771 tumor-bearing mouse models were constructed.The results showed that compared with the paclitaxel or ICOS agonist group,the paclitaxel plus ICOS agonist group had significantly smaller tumors,confirmed that ICOS agonist can develop the ability to kill tumor of paclitaxel on TNBC cell lines.4.ICOS agonists increase the efficacy of chemotherapy by enhancing CD8+T cell functionTo identify the immune cell subsets and ICOS agonist affecting the efficacy of paclitaxel,we performed flow cytometry of the distribution of immune cells after different treatments(Control group;Paclitaxel group;ICOS agonist group;Paclitaxel plus ICOS agonist group).The CD8+T lymphocytes and the ratio of CD8+ICOS+T/CD8+T lymphocytes were increased.The effector subsets of CD8+T cell(TNFa+and Granzyme B+T cell)were enriched.Conclusions1.The predictive model based on DNA mutation sequencing and clinicopathological features can indicate efficacy of NAC and survival with high sensitivity and specificity in breast cancer.2.The dynamic clearance of ctDNA can indicate efficacy of NAC and DFS.The predictive model integrated dynamic ctDNA status with genomic DNA and clinicopathological features has higher sensitivity and specificity.3.ICOS and its immune-related pathways are related to the response to NAC and OS in TNBC patients.ICOS agonist can enhance the anti-tumor effect of chemotherapy by enhancing the effector of CD8+T cells in TNBC.更多还原