Screening anlotinib responders via blood‐based proteomics in non‐small cell lung cancer

Anlotinib has been demonstrated to be effective in advanced non‐small cell lung cancer (NSCLC) patients. The response stratification of anlotinib remains unclear. In this study, plasma samples from 28 anlotinib‐treated NSCLC patients (discovery cohort: 14 responders and 14 non‐responders) were subjected to proteomic analysis, and plasma samples from 35 anlotinib‐treated NSCLC patients (validation cohort) were subjected to validation analysis. Liquid chromatography–tandem mass spectrometry analysis was performed on samples with different time points, namely baseline (BL), best response (BR), and progression disease (PD). Bioinformatics analysis was performed to screen for the underlying differential proteins. Enzyme‐linked immunosorbent assay was performed to detect plasma ARHGDIB, FN1, CDH1, and KNG1 levels respectively. The Kaplan–Meier survival analysis was used for biomarker‐based responsive stratification. Our results indicated that differential proteins between responders and non‐responders showed that proteomic technology potentially contributes to biomarker screening in plasma samples at BL. Furthermore, our results suggested that the detection of plasma ARHGDIB, FN1, CDH1, and KNG1 levels have potential predictive value for anlotinib response both in the discovery cohort and validation cohort. Collectively, this study offers novel insights into the value of plasma biomarker screening via proteomic examination and suggests that plasma ARHGDIB, FN1, CDH1, and KNG1 levels could be used as biomarkers for anlotinib stratification in NSCLC patients.


| BACKGROUND
Anlotinib has been used for third-line or after third-line therapy in advanced non-small cell lung cancer (NSCLC) patients in China. 1,2 Clinical evidence has demonstrated that anlotinib significantly prolongs the progressionfree survival (PFS) (anlotinib vs. placebo; 5.37 vs. 1.40 months) and median overall survival (OS) (anlotinib vs. placebo; 9.63 vs. 6.30 months) at third-line or after third-line therapy. 1 Recent studies have indicated that anlotinib brings survival benefits to NSCLC patients and enhances the survival time of other types of cancers, such as small cell lung cancer (SCLC) and renal cell carcinoma (RCC). [3][4][5] Mechanistically, anlotinib-induced inhibition of angiogenesis and proliferative signaling contributes to the clinical phenomenon. 6,7 Further evidence indicates the targets of anlotinib, such as the chemokine ligand 2 (CCL2), receptor tyrosine kinases vascular endothelial growth (VEGFR), platelet-derived growth factor receptor (PDGFR), and fibroblast growth factor receptor (FGFR) signaling pathways. 8,9 Therefore, the multi-target characteristics of this tyrosine kinase inhibitor (TKI) make clinical biomarker screening difficult.
In recent years, our study suggested that an anlotinibinduced CCL2 decrease could potentially be used as a predictive factor for clinical stratification. 8 Furthermore, blood KLK5 and L1CAM levels have also demonstrated potential value for the screening of anlotinib responders. 10 Subsequently, next-generation sequencing (NGS) for plasma cell-free DNA (cfDNA) plays a role in anlotinib-responsive stratification, and the biomarker tumor mutation index (TMI) plus IDH1 exon4 mutation status can significantly identify anlotinib responders. 11 In addition, other recent studies introduced predictors, including CD31-labeled circulating endothelial cells and baseline characteristics of patients, for the stratification of patients treated with anlotinib. 12,13 Although multiple attempts have been made, the effective biomarker for anlotinib stratification remains unclear. Due to the complex architecture of the anlotinib-induced anti-angiogenic signaling pathway, 14 the importance of investigating plasma samples through a proteomic approach to identify patients who will have a positive response to anlotinib has not yet been justified. Therefore, we performed proteomic analysis on the plasma samples (anlotinib responders and anlotinib non-responders) and sought to examine its predictive value for anlotinib stratification.

| Selection of patients
In the ALTER0303 study (https://clini caltr ials.gov/ NCT02388919), a total of 440 qualified advanced NSCLC patients were enrolled and completed the clinical study. Among the sample libraries, we selected 28 patients as the discovery cohort according to the following three criteria: (1) non-responders (PFS < 60 days) with the best clinical objective response were defined as progression disease (PD); responders (PFS > 80 days) with the best clinical objective response were defined as partial response (Table S1). (2) The plasma samples were collected from the same patient at different time points (baseline (BL), best response (BR), and PD).
(3) All plasma samples passed quality control (including blood collection, centrifugation, plasma collection, and so on). Briefly, the plasma samples from 14 anlotinib non-responders at two time points of BL and PD and the plasma samples from 14 anlotinib responders at three time points: BL, BR, and PD were considered for this study (Figure 1). The clinical characteristics of the 28 NSCLC patients are shown in Tables S1 and S2. All other administration and clinical care information has been introduced in our previous study. 1,2 Furthermore, we picked out 35 patients who were without the limitations of specific clinical characteristics as the validation cohort and collected the plasma samples at BL for biomarker validation.

| Plasma collection and processing
The samples were collected as per standard procedures. Briefly, peripheral blood samples from advanced NSCLC patients were collected using EDTA tubes. All samples were centrifuged (1600 g for 10 min) within 2 h of blood collection. Then, the upper plasma was transferred to 1 ml sterile Eppendorf tubes using a pipette and stored at −80°C. Due to the limitation of plasma volume, mixed plasma was prepared for proteomic analysis in the present study. Seven plasma samples (patients No. 1-7) from responders at BL were mixed together. Each patient contributed 100 μl plasma, and a total of 700 μl plasma was mixed for protein extraction. Similar to the previous procedure, the other seven plasma samples (patients No. [8][9][10][11][12][13][14] from anlotinib, biomarker, liquid biopsy, non-small cell lung cancer, proteomics responders at BL were mixed as duplicates for protein extraction (Tables S1 and S3). Therefore, two mixed samples from responders at BL were used for proteomic analysis. This study evaluated three time points (BL, BR, and PD) for responders and two time points (BL and PD) for non-responders. The mixed samples for other time points were based on the procedures of samples from responders at BL. All samples were collected from the same batch of patients at different time points (Table S3). Ethical approval for this study was obtained from the institutional ethics committee of the Shanghai Chest Hospital prior to commencing. Consent for all resources (including patients' clinical data, blood samples, and tumor samples) was obtained before the clinical screening.

| Protein extraction
Plasma samples of 700 μl were used for protein extraction. Each mixed sample has performed the removal of IgG, IgA, albumin, antitrypsin, haptoglobin, transferrin, and so on using a Thermo Scientific high-select top 14 abundant protein depletion resin kit (Thermo Fisher, USA) ( Table S4). As the previous study introduced, 14 each mixed sample performed standard procedures including suspension (50 μl PBS), centrifugation (10 000 g for 30 min at 4°C), and then suspension using 100 μl lysis buffer (7 M urea, 2 M thiourea). The suspended samples were performed centrifugation at 40 000 g for 30 min. Ultrasonic sonication was used for protein extraction. Protein samples were quantified using the bicinchoninic acid (BCA) method. Gel electrophoresis F I G U R E 1 The flowchart of proteomics of plasma samples from NSCLC patients with anlotinib responders and anlotinib non-responders. Plasma samples were performed for protein extraction, TMT labeling, LC-MS/MS identification, and quantitation for proteomics analysis. BL, the plasma was collected at baseline. BR, the plasma was collected at BR. PD, the plasma was collected at PD.

M/Z
images of the total protein and the total protein without abundant protein were shown in Figure S1.

| Protein digestion
Protein digestion was performed as previously described. 15,16 Briefly, 300 μg of the total protein without abundant protein for each sample were subjected to enzymatic hydrolysis. DTT was added to a concentration of 100 mM, incubated at 100°C for 5 min, and then cooled to room temperature. After precipitating with trichloroacetic acid for 30 min on ice, then diluted with 50 mM NH 4 HCO 3 buffer to a final concentration of 0.5 mg/ml. The samples were performed digestion with trypsin at 37°C for 12 h. Trifluoroacetic acid (TFA) was added to achieve 0.1% TFA in the solution, and desalination was performed using a C18 cartridge.

| Liquid chromatography-tandem mass spectrometry analysis
Peptide mixtures were subjected to nano-liquid chromatography associated with MS for protein identification. MS analysis was performed on the abovementioned 15 components serially. High-performance liquid chromatography (HPLC, Agilent, USA) was used to separate the components. After separation, Q-Extractive Plus MS (Thermo Scientific, USA) was used for MS/MS analysis. Collection of desalting and separating samples were performed using a RP trap column (Thermo EASY column SC200, 150 μm × 100 mm) and a C18 reverse-phase column (Thermo EASY column SC100 traps, 150 μm × 20 mm). Mobile phase A consisted of HPLC-grade water containing 0.1% formic acid (FA), and phase B consisted of 84% HPLC-grade acetonitrile (ACN) containing 0.1% FA. The analytical separation was run at a flow rate of 400 nl/min using a linear gradient of phase B as follows: 0%-45% for 100 min, 45%-100% for 8 min, and 100% for 12 min. Each LC-MS/MS analysis was repeated three times to reduce technical variation. 17

| Differential proteins analysis
This study performed differential proteins comparison among different groups (responders and non-responders) and at different time points (BL, BR, and PD). For the anlotinib responders, we compared the differential proteins between the two time points of BL and BR and compared the differential proteins between the three time points of BL, BR, and PD. For the anlotinib non-responders, we compared the differential proteins between the two time points of BL and PD. Furthermore, we compared the differential proteins between responders and non-responders, at different time points of BL and PD, respectively.

| Bioinformatics analysis
Protein clustering was performed on the different proteins between different samples. Similar to our previous study, 18 Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using a public bioinformatics resource platform (DAVID, https://david.ncifc rf.gov/) by uploading the differential gene lists. Under GO analysis, biological processes, molecular functions, and cellular components were used to characterize the differential proteins.

| Integrative analysis for upregulated proteins and downregulated proteins
We performed the integrative analysis to compare the protein levels of responders and the protein levels of nonresponders at BL. These proteins with significantly higher or lower levels in responder at BL were screened out for further analysis. Next, we compared these high or low levels of proteins to all proteins of responders at BR, and screened out the upregulated proteins and downregulated proteins, respectively. Lastly, we compared the abovescreened proteins to all proteins of responders at PD and screened out the upregulated proteins and downregulated proteins of responders at BL, respectively.

| Detection of plasma protein levels and anlotinib responsive analysis
The enzyme-linked immunosorbent assay (ELISA) kit for ARHGDIB detection was purchased from Cloud-Clone Corporation (CHN). The ELISA kits for FN1, CDH1, and KNG1 were purchased from Abcam (UK). The experimental procedures were performed according to the manufacturer's instructions. Here, we allocated the 28 NSCLC patients (including 14 responders and 14 non-responders) who received anlotinib as a discovery cohort and allocated the 35 NSCLC patients who received anlotinib as a validation cohort. Plasma samples from the discovery cohort and validation cohort were detected at the time point of BL. Based on the levels of ARHGDIB, FN1, CDH1, and KNG1, the binary bit method was used for responsive stratification in the discovery cohort as previously reported. [19][20][21][22] The cutoff values were used for testing the stratification effects in the validation cohort.

| Statistical analysis
Due to the biological repetition being set as two in the present study, the t-test p-value cannot be calculated. Therefore, the significance of differential proteins was calculated according to fold change, without processed statistical adjustment. PFS and OS were summarized as median values and two-sided 95% confidence intervals and were analyzed using the Kaplan-Meier method. The Mantel-Cox test was used to perform the Kaplan-Meier survival analysis using GraphPad Prism 5. Differences were considered significant at p < .05, p < .01, and p < .001.

| Proteomic analysis in anlotinib responders from baseline to progression disease
To analyze the dynamic changes of the differential proteins from BL to PD, we further compared the plasma protein levels of responders at three time points: BL, BR, and PD. A total of 18 proteins were screened and showed  Figure 3A). Biological process analysis suggested that these 18 proteins were enriched in processes such as platelet activation, cell activation, and blood coagulation. Cell component analysis suggested that these proteins were enriched in cell membrane-bound vesicles, extracellular space, and vesicles ( Figure 3B). KEGG pathway analysis indicated that these proteins were enriched in signaling pathways including shigellosis, complement and coagulation cascades, Salmonella infection, and salivary secretion ( Figure 3C). These results suggested that anlotinib-induced plasma protein level alterations at different time points may affect the different biological processes, cell components, and signaling pathways, which are potentially involved in acquired resistance.

| Analysis of primary resistance to anlotinib via proteomic characterization
To screen the primary resistant plasma protein markers of anlotinib via proteomic characterization, we first compared the differential proteins after anlotinib administration among non-responders at two time points, BL and PD.
Compare to BL of non-responders, a total of 41 differential proteins were screened via proteomic analysis in plasma samples from PD of non-responders. Of these 41 proteins, it was found that the level of 20 proteins (COL19A1, TF, B4E1B2 (Protein ID) (Protein ID), KRT2, ITIH1, S100A12, VCL) was remarkably decreased ( Figure 4A). Biological process analysis suggested that these proteins were enriched in processes such as cell migration, cell localization, and cell motility ( Figure 4B). Further analysis indicated that the up-regulated proteins were enriched in metabolic, organic substance metabolic, and cellular processes. The downregulated proteins were enriched in endocytosis, receptor-mediated endocytosis, and chemical homeostasis ( Figure 4C). Cell component

Fold change (R_BL / R_BR) p-value
analysis suggested that these proteins were enriched in extracellular regions and vesicles ( Figure 4B). These results suggested that these proteins may play important roles in tumor cell-activating compensatory effects to relieve anlotinib stress.
Furthermore, the differential proteins between responders and non-responders at BL are still unclear. Here, we identified 470 proteins that existed in the BL samples from responders and non-responders. Compared to the non-responders, 23 proteins (CASP14, CFI,   (Protein ID), PAPOLA, S100A9) were downregulated in responders at BL ( Figure 5A,B). These differential proteins are enriched in biological processes like peptide cross-linking, regulation of peptidase activity, and cornification ( Figure 5C). Cell component analysis suggested that these differential proteins were enriched in the extracellular space and region ( Figure 5C). In particular, these up-regulated proteins in responders were enriched in responses to stimuli, biological regulation, and regulation of biological processes, while the downregulated proteins were enriched in the cellular macromolecule metabolic, nitrogen compound metabolic, and macromolecule metabolic processes ( Figure 5D). Furthermore, the differential proteins between responders and nonresponders at the time of PD suggested that there were 15  proteins with higher levels in non-responders than those in non-responders, and 4 proteins with lower levels in non-responders than those of responders ( Figure S2).

| Integrative analysis reveals bloodbased proteomics potentially be used for screening of anlotinib responders
To further screen out the potential plasma biomarker, integrative analysis was performed on those differential proteins at three time points: BL, BR, and PD from responders and at the time point of BL from non-responders. After the filtered analysis, we found that 43 differential proteins (38 proteins with a low level and 5 proteins with high level) of responders at BL showed important potential values ( Figure 6A). For the samples from nonresponders at BL, of 43 proteins, 5 proteins (COL19A1, KNG1, CF1, RBM42, APOA2) were found to have lower levels, and 38 proteins (ARHGDIB, FN1, CDH1, MENT,  AFTPH, IGLC7, APOE, CLEC3B, KRT2, PAPOLA, CD4,  CTDP1, HYOU1, PDIA3, IGHV3-30 (C) Analysis of biological process, cell component, and molecular function for those differential proteins. *R_BL, the plasma collected from responders at BL. R_BR, the plasma collected from responders at BR. R_PD, the plasma collected from responders at PD. NR_BL, the plasma collected from non-responders at BL.   TPM4, ACTB, SH3BGRL3, YWHAZ, MUC5B,  CAPN1, ALDOA, LCN2, LAMA2, DAG1, PFN1, LTA4H,  TAGLN2, LMAN2, FLNA, ENPP3, S100A6, S100A12, PPIA, SAA1) were found to have higher levels than those of responders at BL (Figure 6A,B). Biological process analysis suggested that these proteins are enriched in receptormediated endocytosis, platelet degranulation, and innate immune response. Cellular component analysis suggested that these proteins are enriched in protein binding, poly (A) RNA binding, and calcium ion binding. Molecular function analysis suggested that these proteins are enriched in extracellular exosome, extracellular region, and space ( Figure 6C). Based on these differential proteins, we detected the plasma ARHGDIB, FN1, CDH1, and KNG1 levels at BL in responders and non-responders, and found that low levels of plasma ARHGDIB, FN1, and CDH1 in NSCLC patients had a better response to anlotinib than those patients with a high level of plasma ARHGDIB, FN1, and CDH1 ( Figures 7A and S3A). On the contrary, the NSCLC patients with a high level of plasma KNG1 had a better response to anlotinib than those patients with a low level of plasma KNG1 (Figures 7A and S3A). To understand whether this phenomenon can be validated in a validation cohort, we further detected the above proteins at BL in a 35 patients' cohort who also received anlotinib therapy ( Figure S3B). Interestingly, the Kaplan-Meier plots analysis suggested that these plasma levels of proteins can also stratify the responders and non-responders in the validation cohort ( Figure 7B). Collectively, these results suggested that proteomics analysis can potentially be used for anlotinib-responsive stratification (Figure 8).

| DISCUSSIONS
Anlotinib, a multi-target TKI inhibitor, has been demonstrated to be effective against different cancers in clinical trials. 1,3,4,5,7 Due to the main mechanism of anti-angiogenesis, the complex architecture of the signaling pathway limits anlotinib biomarker screening. 11 Previous studies have discussed potential anlotinib biomarkers. 8,10,11 However, the effective biomarker for anlotinib stratification remains unclear. In order to screen for potential biomarkers, we performed proteomics on plasma samples from anlotinib responders and anlotinib non-responders.
Proteomic technology has been widely used in biomedical research since its establishment. [23][24][25][26] For example, Jiang et al. found that proteomics could identify novel therapeutic targets for hepatocellular carcinoma 27 ; Eckert et al. suggested that the metabolic regulator NNMT can be screened out via proteomics. 28 Along with technology development, cancer biomarker screening based on proteomics has been introduced in clinical translational research. 25,29,30 Whether proteomic technology can be used for biomarker screening of the anlotinib-induced complex anti-angiogenic signaling pathway is still unclear. Furthermore, the development of liquid biopsy is changing the clinical practice of cancer. 25,31,32,33 Studies have shown the potential value of screening cancer biomarkers from the blood via proteomics. 25,34,35,36,37 However, proteomics-based liquid biopsy-guided cancer treatment still requires further exploration. 25 In the present study, we first compared the changes in plasma protein levels in anlotinib responders between BL and BR and found that proteins with significantly increased or decreased levels may play an important role in anlotinib-induced antitumor effects. After comparing the plasma protein levels at three different time points (BL, BR, and PD) in anlotinib responders, we found that eight differential proteins were potentially associated with anlotinib-acquired resistance. Lastly, after integrating analysis of responders and non-responders at different time points (BL, BR, and PD), the results suggested that 43 differential proteins could potentially be used as biomarkers. Of the 43 differential proteins, ARHGDIB, FN1, CDH1, and KNG1 were demonstrated to be potentially effective for screening anlotinib responders in the present study. However, whether these plasma proteins were derived from cancer cells or normal tissue cells needs to be further discussed. Interestingly, we found the binding proteins including FN1 and CDH1 in the present study as well as the binding protein L1CAM in our previous report 10 potentially played an important role in predicting anlotinib response. Although we found many differential proteins between anlotinib responders and anlotinib nonresponders, the methods for sample preparation (such as mixed samples) still need to be optimized. Furthermore, a significant difference in serval proteins observed between duplicated samples also should be concerned. Whether individual variation or technology bias needs further discussion. The main limitation of the present study was using a mixture of the seven samples as one testing sample. In the future, the proteomic analysis should be performed on individual samples for further validation. Nevertheless, the present study provides an interesting approach for anlotinib biomarker screening via proteomic technology.

| CONCLUSION
In conclusion, our study provides a new perspective on plasma biomarker screening via proteomic examination and suggests that blood-based proteomics could potentially be used as a biomarker for anlotinib stratification in NSCLC patients.