第一部分：免疫检査点分子PD-L1表达在头颈部鳞状细胞癌的临床病理意义

背景：程序性细胞死亡受体配体-1（Programmed cell death ligand 1，PD-L1）在头颈部鳞状细胞癌（Head and neck squamous cell carcinoma，HNSCC）中的表达被用作预测免疫治疗疗效的生物标志物，综合阳性评分（Combined positive score，CPS）被FDA推荐用于HNSCC患者PD-L1表达的评分标准。

目的：目前很少有以CPS为标准的关于HNSCC中PD-L1表达的研究，同时这些已有的研究结果相互矛盾，PD-L1的阳性表达率与相关临床因素仍存在争议。因此，用CPS的评估方法准确评价HNSCC中PD-L1的表达仍有价值。

方法：我们回顾性收集了119例在中国医学科学院肿瘤医院确诊并接受手术的原发性HNSCC患者，并排除了术前接受化疗或放疗的患者。119例HNSCC患者中包括34例口腔鳞状细胞癌（Oral cavity squamous cell carcinomas，OSCC）、9例口咽鳞状细胞癌（Oropharyngeal squamous cell carcinomas，OPSC）、36例下咽鳞状细胞癌（Hypopharyngeal squamous cell carcinomas，HPSC）和40例喉鳞状细胞癌（Laryngeal squamous cell carcinomas，LSC）。我们采用单克隆抗体22C3（Dako）进行免疫组织化学染色，CPS评分来评估PD-L1的表达，并分析PD-L1表达与临床病理特征的关系。PD-L1表达与临床病理参数之间的相关性采用卡方检验或Fisher 's精确检验。

结果：

1) 当CPS截断值设置为1时，107例HNSCC患者（89.9%）PD-L1表达呈阳性。当CPS截断值设置为20时，52例HNSCC患者（43.7%）PD-L1表达呈阳性。

2) 肿瘤分期（P = 0.022）、肿瘤部位（P = 0.004）与PD-L1表达相关（CPS≥1），非糖尿病患者（P = 0.046）与PD-L1阳性表达相关（CPS≥20）。无论截断值是1还是20，PD-L1表达与年龄均无显著相关性（P = 0.083，P = 0.085）。

3) 对于40个LSC样本，当CPS截断值设置为20时，我们发现N分期与PD-L1阳性表达显著相关（P = 0.042）。

4) 在评估36个HPSC样本中PD-L1表达时，PD-L1阳性表达（CPS≥1）与年轻患者（P = 0.030）和无II型糖尿病患者（P = 0.040）有关。当CPS截断值设置为20时，T1期（P = 0.024）与PD-L1阳性表达相关。

5) 单因素Cox回归分析显示临床特征与总生存期（Overall survival，OS）或无进展生存期（Progression-free survival，PFS）无显著相关性。临床预后因素采用Log-rank检验进行独立评估，无论CPS的截断值设置为1还是20，PD-L1的表达与OS或PFS均无相关性。

结论：我们用22C3单抗回顾性检测HNSCC患者PD-L1的表达，并研究了PD-L1表达与临床病理参数之间的关系，这些患者更有可能受益于抗PD-1/PD- L1治疗。在我们的研究中，伴II型糖尿病的HNSCC与PD-L1的表达呈负相关，这可能与II型糖尿病涉及的免疫机制有关，免疫治疗在II型糖尿病患者中的效果值得进一步探讨。

第二部分：基于单细胞和批量RNA测序数据建立并验证头颈部鳞状细胞癌预后预测的B细胞标记基因模型

背景：虽然经典的肿瘤TNM分期已被广泛应用于临床诊断并协助临床治疗决策，但头颈部鳞状细胞癌（Head and neck squamous cell carcinoma，HNSCC）作为一种高度异质性的恶性肿瘤，TNM分期的定义仅基于肿瘤本身，未能准确预测HNSCC患者的预后。越来越多的研究表明，肿瘤微环境中的各种免疫细胞尤其是B细胞在肿瘤发生发展和预后中起着至关重要的作用。

目的：本研究中我们旨在建立一种B细胞标记基因相关的预后模型，以对HNSCC患者的预后预测提供指示意义。

方法：我们从GEO数据库中获得了GSE164690的18份初治HNSCC样品的单细胞RNA测序数据（single-cell RNA-sequencing，scRNA-seq），并从中分析了HNSCC的B细胞标记基因（B cell marker genes，BCMG）。从UCSC-XENA下载499例TCGA-HNSCC原发灶的转录组表达谱数据作为训练组，从GEO数据库和ArrayExpress数据库获得了四个具有临床信息和随访信息的转录组表达数据集作为验证组来验证预后模型的效能，包括GSE41613、GSE42743、GSE65858和E-MTAB-8588表达谱，分别包含97、74、270和108个HNSCC样本。我们使用单因素Cox回归，LASSO回归及多因素Cox回归模型的方法建立了预后预测风险模型 。根据风险指数评分，将患者分为高风险组和低风险组。基于多因素Cox回归分析获得了独立因子并构建了TCGA-HNSCC患者的2年、3年和5年OS的列线图。为进一步解释目标基因的生物学功能，我们进行了基因本体论（GO，gene ontology）富集分析及京都基因与基因组百科全书（Kyoto Encyclopedia of Genes and Genomes，KEGG）途径分析。此外，我们利用表达数据ESTIMATE算法分析了TCGA-HNSCC中基质细胞和免疫细胞的分数，并通过CIBERSORT算法估计了高风险和低风险人群中免疫细胞浸润的丰度，分析了风险评分与肿瘤微环境的关系。我们使用GEPIA2分析了TCGA-HNSCC肿瘤组织与匹配正常组织之间的表达差异，数据来自TCGA正常和GTEx数据集。我们收集了在中国医学科学院肿瘤医院确诊的原发性HNSCC的114例HNSCC组织，并利用福尔马林固定石蜡包埋（Formalin-fixed paraffin-embedded，FFPE）标本构建组织微阵列（Tissue microarrays，TMA）。我们利用Ventana BenchMark平台的自动化系统进行了免疫组织化学（Immunohistochemistry，IHC）染色，对模型中的基因和一些免疫检查点分子编码的蛋白质进行染色，包括FTH1，PD-1，CTLA-4，TIM-3，LAG-3，TIGIT，PD-L2及PD-L1。

结果：

1) 我们通过对GEO数据库中HNSCC的scRNA-seq数据进行综合分析，筛选了145个BCMG。

2) 我们鉴定出了七个基因（FCRLA，FTH1，LAT，S100A4，TMP1，TNFRSF18和TXN）用来建立B细胞标记基因模型（BCMG signature，BCMGS)。

3) 利用TCGA-HNSCC数据集建立了BCMGS，并在四个独立的数据集中进行了验证。

4) 多因素Cox回归分析将该模型确定为独立的预后因素，我们利用该模型及其他独立风险因素构建预后列线图。

5) 对免疫谱的研究表明，低风险组的患者表现出更丰富的免疫细胞浸润。此外，低风险组的特点是TCR和BCR多样性较高，这表明低风险组患者可能对免疫治疗更敏感。

6) 我们使用GEPIA2分析了模型中的7个基因在TCGA-HNSCC肿瘤组织与匹配正常组织之间的表达差异。我们发现与正常组织相比，FTH1和TNFRSF18的蛋白表达在肿瘤组织中显著高表达。而FCRLA、LAT、S100A4、TMP1和TXN的蛋白表达在肿瘤组织和正常组织之间没有发现显著相关性。

7) 进行免疫组织化学染色，发现FTH1高表达与OS不良显著相关（P = 0.025）。TIM-3、LAG-3和PD-1的表达与HNSCC患者较好的OS相关。然而，PD-L1、PD-L2、CTLA-4、TIGIT和预后之间没有统计学意义差异。

结论：BCMGS是HNSCC的一种有希望的预后生物标志物，可能有助于解释免疫治疗的反应，并为未来HNSCC治疗的研究提供新的视角。

Part one: The clinicopathological significance of PD-L1 expression in head and neck squamous cell carcinoma
Background: Programmed cell death ligand 1 (PD-L1) expressed in head and neck squamous cell carcinoma (HNSCC) was used as a predictive biomarker for immunotherapy, and FDA recommended the combined positive score (CPS) for PD-L1 scoring in patients with HNSCC.
Objective: There are few studies on the expression of PD-L1 in HNSCS using CPS as the standard. Meanwhile, the results of these existing studies are contradictory, and the positive expression rate of PD-L1 differs from related clinical factors. Therefore, it is still valuable to accurately analyze the expression of PD-L1 in HNSCS by the CPS assessment method.
Methods: We retrospectively collected 119 patients with primary HNSCC diagnosed and operated in the Chinese Academy of Medical Sciences Cancer Hospital. Including 34 cases of oral cavity squamous cell carcinomas (OSCC), 9 cases of oropharyngeal squamous cell carcinomas (oropharyngeal squamous cell carcinomas, OPSC), 36 hypopharyngeal squamous cell carcinomas (HPSC), and 40 cases of laryngeal squamous cell carcinomas (LSC). Patients who received chemotherapy or radiation before surgery were included. Monoclonal antibody 22C3 (Dako) was used for immunohistochemical staining and CPS score to evaluate the expression of PD-L1. And we analyzed the relationship between PD-L1 expression and clinicopathological features by Chi-square test or Fisher's exact test.
Results:
1) When the CPS threshold was set to 1, the 107 HNSCC patients (89.9%) were positive for PD-L1 expression. PD-L1 expression was positive in 52 HNSCC patients (43.7%) when the CPS cutoff was set to 20.
2) Tumor stage (P = 0.022) and tumor site (P = 0.004) were correlated with PD-L1 expression (CPS≥1), and non-diabetic patients (P = 0.046) were correlated with positive PD-L1 expression (CPS≥20). Whether the cutoff value was 1 or 20, there was no significant correlation between PD-L1 expression and age (P = 0.083, P = 0.085).

3) For 40 LSC samples, when the CPS threshold was set at 20, N stage was significantly correlated with the positive expression of PD-L1 (P = 0.042).
4) When assessing PD-L1 expression in 36 HPSC samples, positive expression of PD-L1 (CPS≥1) was associated with young adults (P = 0.030) and patients without type 2 diabetes (P = 0.040). When the CPS threshold was set to 20, the T1 stage (P = 0.024) was associated with positive expression of PD-L1.
5) Univariate Cox regression analysis showed no significant correlation between clinical features and OS or PFS. Clinical prognostic factors were assessed independently by a Log-rank test. However, no matter the threshold of CPS was set to 1 or 20, PD-L1 expression was not correlated with OS or PFS.
Conclusions: We used 22C3 retrospectively to detect the expression of PD-L1 in HNSCC patients, and studied the relationship between PD-L1 expression and clinicopathological parameters in patients more likely to benefit from anti-PD-1 / PD-L1 therapy. In our study, HNSCC with type II diabetes were negatively correlated with the expression of PD-L1, which may be related to the immune mechanism involved in type II diabetes. The effect of immunotherapy in patients with type II diabetes deserves further discussion.
 

Part two: Comprehensive analysis of single-cell and bulk RNA-sequencing data identifies and verifies B cell marker genes signature that predicts prognosis in head and neck squamous cell carcinoma
Background: Although the classic TNM staging of head and neck squamous cell carcinoma (HNSCC) has been widely used in clinical diagnosis and to assist clinical treatment decisions, the definition of TNM staging of HNSCC is based solely on the tumor itself as a highly heterogeneous malignancy. It is not enough to accurately predict the prognosis of HNSCC patients. More and more studies have shown that various immune cells in the tumor microenvironment, especially B cells, play a crucial role in tumor development and prognosis.
Objective: In this study, we aimed to develop a B-cell marker gene-related model to indicate the prognosis of HNSCC patients.
Methods: We obtained 18 primary HNSCC samples of GSE164690 from the GEO database single-cell RNA-sequencing (RNA-Seq) and analyzed B cell marker genes of HNSCS. Transcriptome expression profile data of 499 TCGA-HNSCC primary sites were downloaded from UCSC-XENA as a test group. Four transcriptome expression data sets with clinical annotation and follow-up information were obtained from GEO and ArrayExpress databases as validation sets to verify the efficacy of the prognostic model. The expression profiles of GSE41613, GSE42743, GSE65858 and E-MTAB-8588 contained 97, 74, 270 and 108 HNSCC samples, respectively. We used univariate Cox regression, LASSO regression, and multivariate Cox regression to establish a prognostic risk model. According to the risk index score, patients were divided into high-risk and low-risk groups. Based on the independent factors obtained by multivariate Cox regression analysis, the 2-year, 3-year and 5-year OS charts of TCGA-HNSCC patients were constructed. To further explain the biological functions of the target Genes, GO (gene ontology) analysis were conducted and the KEGG (Kyoto Encyclopedia of Genes and Genomes) approach was analyzed jointly. In addition, we analyzed stromal and immune cell scores in TCGA-HNSCC using the ESTIMATE algorithm, estimated the abundance of immune cell infiltration in high- and low-risk groups using the CIBERSORT algorithm, and analyzed the relationship between risk scores and tumor microenvironment. Expression differences between TCGA-HNSCC tumor tissues and matched normal tissues were analyzed using GEPIA2 from the TCGA normal and GTEx datasets. We collected tissue from 114 primary HNSCC diagnosed in the Cancer Hospital of Chinese Academy of Medical Sciences. Tissue microarrays (TMAs) were constructed from formalin-fixed paraffin-embedded (FFPE) samples. Immunohistochemistry (IHC) was performed by the automated system of Ventana BenchMark platform. Genes in the model and proteins encoded by immune checkpoints were stained, including FTH1, PD-1, CTLA-4, TIM-3, LAG-3, TIGIT, PD-L2 and PD-L1.
Results:
1) We comprehensively analyzed the Single cell RNA-Seqing data of HNSCC in GEO database and screened 145 B cell marker genes (BCMG).
2) We identified seven genes (FCRLA, FTH1, LAT, S100A4, TMP1, TNFRSF18, and TXN) to establish B cell marker gene signature (BCMGS).
3) BCMGS were established using the TCGA-HNSCC dataset and verified in four independent datasets.
4) Multivariate Cox regression analysis identified this model as an independent prognostic factor, and we used this model to construct a prognostic nomogram with other clinical factors.
5) Studies of immune profiles have shown that patients in the low-risk group exhibit immune cell infiltration. In addition, the low-risk group was characterized by a higher diversity of TCR and BCR, suggesting that low-risk patients may be more sensitive to immunotherapy.
6) We assessed the expression of seven genes in BCMGS between TCGA-HNSCC tumor tissues and matched normal tissues using GEPIA2. The protein expression of FTH1 and TNFRSF18 was significantly highly expressed in tumor tissues compared with normal tissues. No significant correlation was found in the protein expression of FCRLA, LAT, S100A4, TMP1, and TXN between tumor tissues and normal tissues.
7) Immunohistochemical examination showed that high expression of FTH1 was significantly correlated with OS difference (P = 0.025). The expressions of TIM-3, LAG-3, and PD-1 were associated with better OS in HNSCC patients. However, there was no statistically significant difference between PD-L1, PD-L2, CTLA-4, TIGIT, and prognosis.
Conclusion: BCMGS is a promising prognostic biomarker in HNSCC that may help explain immunotherapy responses and provide a new perspective for future studies of HNSCC.