introduction
In the past decade, immune checkpoint inhibitors (ICls) have been gradually applied to clinical cancer treatment[1]. ICls target programmed cell death protein 1 (PD-1) on T cells [2], cytotoxic T lymphocyte-associated protein 4 (CTLA4) [3], or PD-L1 on the surface of tumor cells, thereby enhancing anti-tumor immune responses. ICls treatment significantly improved the therapeutic effect of tumors, and bispecific antibodies (DBTs) blocked two immune checkpoint signals at the same time, which had stronger anti-tumor activity than PD-1/PD-L1 inhibitor treatment. However, there has been no accepted predictor of efficacy in immunotherapy, so it is particularly important to explore appropriate predictor markers for efficacy. On May 8, 2024, the team of Professor Ding Chen of the Institute of Human Phenome of Fudan University and the team of Professor Jia Youchao of the Department of Medical Oncology of the Affiliated Hospital of Hebei University published an online paper entitled "Plasma proteome profiling reveals dynamic of cholesterol marker after dual blocker" in the top international academic journal Nature Communications (IF 17.7). therapy", a longitudinal plasma proteomic analysis of QL1706, a PD-1/CTLA-4 dual immune checkpoint blocker, found that cholesterol metabolism was activated in the non-progressive group (DNP) and the biomarker APOC3 was highly correlated with the activation of HDL partial remodeling in the DNP group, and proposed that PA, LDH and APOC3 could be used as potential predictive biomarkers for DBT efficacy.
The progress in the treatment of ICIs is obvious to all, but it must be acknowledged that only a small percentage of patients with advanced tumors can respond to ICIs treatment. Studies have shown that combination blockade therapy with DBTs, such as anti-PD-1 and anti-CTLA-4 ICIs, significantly increases the objective response rate (ORR) in melanoma (from 43.7% to 57.6%) [4] and non-small cell lung cancer (NSCLC) (35.9%) [5]. QL1706 is a PD-1/CTLA-4 dual immune checkpoint blocker that has shown promising antitumor activity in a variety of solid tumors. In a published phase I/Ib trial, the ORR and median duration of response were 16.9 percent (79/468) and 11.7 months, respectively [6]. However, no predictive biomarkers have been developed to identify whether patients will respond to DBT or not. The team of Professor Ding Chen and Professor Jia Youchao has long been committed to the collaborative research of multi-omics combined analysis of tumor biomarkers. In this study, 113 longitudinal plasma samples from 22 cancer patients treated with DBT (QL1706) were included, including 6 cases of LC (lung cancer), 4 cases of CCA (cholangiocarcinoma), 3 cases of RCC (renal cell carcinoma), 3 cases of OVCA (cancer), 2 cases of CRCA (cancer), 2 cases of CESC (cervical squamous cell carcinoma), 1 case of BLCA (bladder cancer), and 1 case of UCEC (endometrial cancer). An independent validation cohort consisting of 54 longitudinal plasma samples from 27 patients receiving anti-PD-1 antibody monotherapy was also collected. In addition, the investigators recruited 24 healthy controls to establish a baseline reference for the plasma proteome. In the end, a total of 191 samples participated in the study (Figure 1). By applying an advanced liquid chromatography DIA-MS pipeline, the researchers generated a large plasma proteome dataset, which combined with the patient's blood routine, blood biochemistry, coagulation function, cardiac enzyme profile, thyroid function, pituitary-adrenal axis, virology, demographics, and clinical pathology, and found that high-abundance proteins were involved in cholesterol metabolism and inflammatory processes, while low-abundance proteins were enriched in response markers of stress and phagocytosis (Fig. 1).
Figure 1. Cohort Characteristics and Study Design (Credit: Nature Communications) In order to further explore the correlation between cholesterol metabolism and response to DBT treatment, a comparative analysis of proteomic data from healthy individuals, before DBT treatment, and after one cycle of DBT treatment found that the proteins in the HLH cluster imply the restoration of tumor suppressive protein expression after DBT, and are involved in lipid binding and antioxidant activity (APOM, CETP, and ALB). The LHL protein cluster is characterized by an extracellular space (HRG, HGFAC), indicating the "calming" of tumor-associated proteins after DBT. The LMH cluster showed significant enrichment of immune-related processes such as lymphocyte-mediated immunity and complement cascades (SAA1, FGA, and CD14) (Figure 2). These results represent a strong effect of DBT on cholesterol processes, immune responses, and oncogenic signaling. At the same time, we assessed the levels of liver function markers in samples from the pre-treatment and first DBT cycle groups. Aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), and γ-glutamyltransferase (GGT) levels reflecting liver impairment were elevated in paired samples. These results suggest that hepatic impairment should be considered in clinical trials of DBT, and the association between DBT and liver function should be further studied.
Figure 2. The plasma proteome after DBT showed a negative correlation between cholesterol metabolism and oncogenic signaling (Credit: Nature Communications), and the researchers then studied the differences in plasma proteome profiles between the DNP and DP groups. By comparing the ssGSEA scores between the two groups, it was found that lipid-related processes and neuron-related processes were upregulated in the DNP group (Figure 3). Lipid-related biological processes upregulated by the DNP group include triglyceride-rich lipoprotein particle remodeling, reverse cholesterol transport, lipoprotein lipase activity, lipoprotein particle clearance, and plasma lipoprotein assembly (Figure 3). Taking into account the effect of BMI on lipid metabolism, the investigators tested BMI between the two groups and observed no significant differences. The levels of high-density lipoprotein components APOC3, APOC2, and APOL1 were higher in the DNP group than in the DP group (permutation-based t-test p=2.00E-4, 1.00E-3, and 7.70E-3) (Figure 3). These results suggest alterations in apolipoprotein in DBT. The above analysis revealed the effect of DBT on APOC3, APOC2, and APOL1 levels, implying the potential role of these proteins in monitoring DBT reactions. In order to screen out valuable proteins, the investigators proposed a criterion to explore biomarkers, for patients without DPs samples, (1) the expression level of the biomarker gradually increased with the treatment cycle, and (2) p<0.05 in the linear model. The results showed that APOC3 and APOC2 met the criteria (Figure 3). In addition, there was a more significant difference in the expression level of APOC3 than APOC2 between the DNP group and the DP group. Therefore, APOC3 was selected as a potential biomarker indicative of DBT response status. At the same time, based on linear regression of time series, it was found that APOC3 levels increased in patients without DP samples, but not in patients with DP samples (Fig. 3). To further validate the potential of APOC3 as a biomarker of DBT response, the investigators also evaluated medical radiological imaging in these patients to fully analyze the relationship between treatment effect and APOC3 protein levels, and found a negative correlation between tumor size and APOC3 protein levels in patients, and also observed the emergence of some new metastases (Figure 3). These results further demonstrate the potential of APOC3 as a biomarker of DBT therapeutic response.
Figure 3. Comprehensive analysis of proteomic and radiological imaging reveals the potential of APOC3 as a biomarker in DBT response (Credit: Nature Communications) Next, the investigators built a machine learning framework based on the clinical (PA, LDH) and proteomic (APOC3) features present in the longitudinal cohort associated with DBT response (Fig. 4) to integrate the features into a predictive model of DBT response to predict the disease progression status of patients. Based on clinical and proteomic features, there are five different feature combinations to choose from, including (1) PA, (2) LDH, (3) APOC3, (4) comprehensive clinical features (PA and LDH), and (5) comprehensive all features (PA, LDH, and APOC3) (Figure 4). Model testing found that the two models, characterized by APOC3 and combining all features, showed better performance than other models with clinical indicators (Figure 4). We evaluated the performance of the model with a combined range of features of 0.87, 0.85, 0.79, 0.85, 0.85, and 0.73 for AUROC, equilibrium accuracy, F1, recall, and accuracy, respectively. At the same time, the investigators found an exciting result when they tested the model in the anti-PD1 monotherapy cohort samples, the model that combined all characteristics showed better performance, followed by the model that integrated clinical characteristics, and in addition, we observed that the APOC3 protein levels in patients with 6 sample points gradually increased during the anti-PD1 monotherapy period. This suggests that PA, LDH, and APOC3 also have the potential to be predictive biomarkers of response to anti-PD1 monotherapy.
图4. 基于蛋白质组临床特征的机器学习模型为DBT提供了准确的测量方法(Credit: Nature Communications)
In summary, the researchers optimized the plasma proteome analysis pipeline and provided a proteomic dataset for the DBT cohort. By integrating clinical and proteomic data, key candidate biomarkers PA, LDH, and APOC3 were discovered. Machine learning models based on proteomic clinical features provide accurate predictions of DBT cohorts. Notably, the predictive power of this model can also be extended to anti-PD-1 treatment cohorts. At the same time, these clinical and proteomic biomarkers explored by longitudinal exploration can also indicate a patient's long-term response to DBT. Finally, this study expands our understanding of the biology of DBT plasma proteins and generates hypotheses that may serve as the basis for future clinical trials of precision immunotherapy response.
bibliography
1. Sharpe, A. H. & Pauken, K. E. The diverse functions of the PD1 inhibitory pathway. Nat. Rev. Immunol. 18, 153–167 (2018).
2. Brahmer, J. R. et al. Safety and activity of Anti–PD-L1 antibody in patients with advanced cancer. N. Engl. J. Med. 366, 2455–2465 (2012).
3. Lipson, E. J. & Drake, C. G. Ipilimumab: an Anti-CTLA-4 antibody for metastatic melanoma. Clin. Cancer Res. 17, 6958–6962 (2011).
4. Larkin, J. et al. Combined Nivolumab and Ipilimumab or Monotherapy in Untreated Melanoma. N. Engl. J. Med. 373, 23–34 (2015).
5. Hellmann, M. D. et al. Nivolumab plus Ipilimumab in advanced non–small-cell lung cancer. N. Engl. J. Med. 381, 2020–2031 (2019).
6. Zhao, Y. et al. First-in-human phase I/Ib study of QL1706 (PSB205), a bifunctional PD1/CTLA4 dual blocker, in patients with advanced solid tumors. J. Hematol. Oncol. 16, 50 (2023).
Original link https://doi.org/10.1038/s41467-024-47835-y
Editor-in-charge|Explore Jun
Typography |
文章来源|“ iNature”
End