Abstract

   EGFR‐mutated lung cancer patients sometimes display restricted
   responses to third‐generation tyrosine kinase inhibitors (TKIs),
   potentially attributable to undervalued input from stromal cells,
   notably pericytes (PCs). The study shows that PCs isolated from
   EGFR‐mutated patients have a unique secretome profile, notably
   secreting IL32 and affecting signaling pathways and biological
   processes linked to TKI sensitivity. Clinical evidence, supported by
   single‐cell RNA sequencing and multiplex immunostaining of tumor
   tissues, confirms the presence of IL32‐expressing pericytes closely
   interacting with β5‐integrin‐expressing cancer cells in EGFR‐mutated
   patients, impacting therapeutic response and prognosis. Co‐culture and
   conditioned medium experiments demonstrate that PCs reduce TKI
   effectiveness in EGFR‐mutated cancer cells, a reversible phenomenon
   through silencing IL32 expression in PCs or depleting the IL32 receptor
   β5‐integrin on cancer cells, thereby restoring cancer cell sensitivity.
   Mechanistically, it is shown that YY1 signaling upregulates IL32
   secretion in PCs, subsequently activating the β5‐integrin‐Src‐Akt
   pathway in EGFR‐mutated cancer cells, contributing to their TKI
   sensitivity. In animal studies, co‐injection of cancer cells with PCs
   compromises TKI effectiveness, independently of blood vessel functions,
   while inhibition of β5‐integrin restores tumor cell sensitivity.
   Overall, the findings highlight direct crosstalk between cancer cells
   and pericytes, impacting TKI sensitivity via IL32‐β5‐integrin paracrine
   signaling, proposing an enhanced therapeutic approach for EGFR‐mutated
   patients.

   Keywords: β5‐integrin, non‐small cell lung cancer, pericytes, tyrosine
   kinase inhibitor sensitivity
     __________________________________________________________________

   This study investigates how pericytes regulate the sensitivity of
   EGFR‐mutated cancer cells to tyrosine kinase inhibitors (TKIs). It is
   found that YY1 transcriptionally upregulates IL‐32 secretion in
   pericytes, which binds to and activates β5‐integrin in cancer cells,
   triggering the Src‐Akt signaling pathway and reducing TKI sensitivity.
   Targeting pericytes’ paracrine effect could restore TKI sensitivity in
   lung cancer.

   graphic file with name ADVS-11-2405130-g006.jpg

1. Introduction

   Lung cancer continues to be the primary cause of cancer‐related deaths
   worldwide, with patients in advanced stages of the disease facing
   limited treatment choices.^[ [44]^1 ^] Non‐small cell lung cancer
   (NSCLC) accounts for ≈85% of lung cancer cases, presenting a
   significant challenge due to our incomplete understanding of its
   underlying pathological mechanisms.^[ [45]^2 ^] Recent advancements in
   next‐generation sequencing have revealed that epidermal growth factor
   receptor (EGFR) mutations are present in ≈10–35% of NSCLC cases, with
   some variations based on region and ethnicity.^[ [46]^3 ^] EGFR, a
   transmembrane receptor tyrosine kinase with a central role in governing
   cancer cell proliferation and survival pathways,^[ [47]^4 ^] has become
   a pivotal therapeutic target for the management of NSCLC tumors,
   potentially surpassing chemotherapy.^[ [48]^5 ^]

   Considerable progress has been made in developing and clinically
   evaluating EGFR tyrosine kinase inhibitors (TKIs).^[ [49]^3 , [50]^6 ^]
   While first‐ and second‐generation TKIs have shown efficacy in treating
   EGFR‐mutated NSCLC,^[ [51]^7 ^] challenges remain due to treatment
   insensitivity from the emergence of the EGFR T790M mutation and
   toxicity issues associated with second‐generation TKIs.^[ [52]^8 ^]
   Third‐generation TKIs,^[ [53]^8a ^] like Almonertinib and Osimertinib,
   have been developed to target EGFR sensitizing and T790M mutations
   while sparing wild‐type EGFR to reduce toxicity.^[ [54]^5 , [55]^9 ^]
   Additionally, recent studies have shown that third‐generation TKIs are
   effective in treating advanced NSCLC patients with EGFR mutations,
   leading to their approval as a first‐line therapy for this condition.^[
   [56]^10 ^] However, despite these drugs having been approved for
   treating NSCLC, some patients still exhibit poor responsiveness,^[
   [57]^11 ^] with mechanisms not fully understood.

   Previous studies on TKI sensitivity have often overlooked the role of
   the tumor microenvironment, especially stromal cells, in influencing
   drug response.^[ [58]^11b ^] This oversight may contribute to the
   limited success of third‐generation TKIs. NSCLC, characterized by
   aberrant vascularization, fosters a supportive tumor microenvironment
   promoting cancer progression.^[ [59]^12 ^] Traditional approaches
   targeting tumor vasculature have had limited clinical success,^[
   [60]^13 ^] potentially worsening tumor hypoxia and reducing
   chemotherapeutic delivery,^[ [61]^14 ^] leading to chemoresistance and
   metastasis.^[ [62]^15 ^] Recent research suggests direct communication
   between pericytes (PCs) and cancer cells through paracrine signaling,
   impacting tumor growth and chemotherapy response.^[ [63]^12 , [64]^16
   ^] However, the influence of PC‐derived signals on EGFR‐targeted
   therapy is unknown.

   Pericytes, acknowledged for their role in regulating inflammatory
   responses, particularly during infections, possess the capability to
   secrete essential proinflammatory factors, which play a vital role in
   facilitating these crucial processes.^[ [65]^16 , [66]^17 ^] Recent
   research has indicated that these proinflammatory factors also have a
   significant impact on propelling cancer growth and progression.^[
   [67]^18 ^] Among these factors, interleukin 32 (IL32), a well‐known
   proinflammatory cytokine, has emerged as a substantial contributor.^[
   [68]^19 ^] IL32 is generated within the tumor microenvironment and has
   been associated with various aspects of cancer progression.^[ [69]^20
   ^] Nevertheless, the precise role of IL32 in either promoting or
   impeding lung cancer cell progression is a topic of ongoing debate. An
   intriguing aspect is the frequent expression of IL32's known receptor,
   arginylglycylaspartic acid (RGD) binding integrin (such as
   β5‐integrin),^[ [70]^21 ^] in lung cancer cells.^[ [71]^22 ^]
   Furthermore, there is an active area of research investigating whether
   IL32 influences cancer cell sensitivity to TKIs and whether PCs are the
   main source of this cytokine for this specific regulatory purpose.

   This study delves into the paracrine influence of pericytes on TKI
   sensitivity in EGFR‐mutated NSCLC. We found that PCs significantly
   increase IL32 secretion, which correlates with adverse outcomes in EGFR
   mutant NSCLC patients. Functional assays highlight PCs' role in
   regulating cancer cell sensitivity to TKIs, mitigated by depleting IL32
   or blocking its receptor β5‐integrin. Mechanistically, YY1 regulates
   the expression and secretion of IL32 in PCs. IL32 then activates the
   β5‐integrin‐Src‐Akt pathway, which impairs TKI effectiveness in
   EGFR‐mutated cancer cells. Co‐injection of cancer cells with PCs
   reduces tumor responsiveness to TKIs, with IL32/β5‐integrin depletion
   restoring sensitivity. This underscores the critical role of
   PC‐secreted IL32 in impacting TKI sensitivity in EGFR mutant NSCLC.

2. Results

2.1. Pericyte‐IL32 Expression Correlates with Patient Prognosis in
EGFR‐Mutated NSCLC Patients

   To identify the paracrine factors predominantly secreted by pericytes
   (PCs) derived from EGFR‐mutated NSCLC patient tumors, we conducted a
   comparative secretomics analysis of conditioned medium (CM) from PCs
   derived from three different EGFR‐mutated patients and the EGFR‐mutated
   NSCLC cancer cell line HCC827. We also compared CM from primary cancer
   cells with CM from paired PCs isolated from the same three EGFR‐mutated
   patients (Figures [72]1A–C; Figures [73]S1A–C, Supporting information).
   Heatmap analysis of the secretomics data unveiled a distinctive
   secretome profile in PCs, characterized by altered secretion patterns
   of cytokines, such as proinflammatory factors IL32 and IL17 family
   members (Figure [74]1A; Figure [75]S1A, Supporting information). Both
   KEGG and GO biological process analyses of the secretomics data
   indicated an enrichment in signaling pathways and biological processes
   related to paracrine signaling in PCs, such as PI3K‐Akt signaling
   pathway, cytokine‐cytokine receptor interaction, EGFR tyrosine kinase
   inhibitor resistance, ECM‐receptor interaction, cell‐cell signaling,
   response to drug and regulation of secretion (Figures [76]1B,C; Figures
   [77]S1B,C, Supporting Information). To validate our findings, cytokine
   proteome XL arrays were performed with PCs and HCC827 cells,
   demonstrating the increased expression of IL22, IL32, CCL5, IL17A, and
   CXCL12 in PCs (Figure [78]1D). Further RT‐PCR analysis consistently
   indicated that IL32 was the only cytokine consistently up‐regulated in
   PCs when compared to EGFR‐mutated cancer cell lines or their paired
   primary cancer cells (Figure [79]1E; Figure [80]S1D, Supporting
   Information). We also conducted a western blot analysis of IL32
   expression in PCs derived from three different EGFR‐mutated NSCLC
   patients, as compared to cancer cell lines, confirming the
   up‐regulation of IL32 in all PCs (Figure [81]1F). To determine the
   specificity of IL32 up‐regulation in PCs, we performed IL32 ELISA
   assays on conditioned medium (CM) harvested from normal fibroblasts
   (NFs), human umbilical vein endothelial cells (ECs), EGFR mutated lung
   cancer cell lines HCC827, PC9 and NCI‐H1975 (T790M+ cells), and primary
   PCs, revealing a significant increase in the secretion of IL32 in PC‐CM
   compared to that from NFs, ECs, or cancer cells (Figure [82]1G). To
   further confirm our observation, we also performed RT‐PCR and ELISA
   analysis with primary cancer cells and paired PCs and their conditioned
   medium, indicating an up‐regulation of IL32 expression in PCs isolated
   from 3 different EGFR‐mutated patient‐derived tumors as compared to
   their paired primary cancer cells (Figures [83]S1D,E, Supporting
   Information). Further, ELISA assays showed an elevated IL32 level in CM
   from PCs isolated from EGFR‐mutated patients as compared to CM from PCs
   isolated from EGFR wild‐type patients (Figure [84]S1F, Supporting
   Information). These findings provide strong evidence that IL32, an
   interleukin (IL) associated with tumorigenesis,^[ [85]^20a ^] primarily
   emanates from PCs within the tumor microenvironment of EGFR‐mutated
   NSCLC patients.

Figure 1.

   Figure 1
   [86]Open in a new tab

   Pericyte‐IL32 expression is associated with cancer progression in NSCLC
   patients harboring EGFR mutations. A) Visualizing a heatmap comparing
   secretome analysis between EGFR mutant cancer cell line HCC827 and PCs
   derived from 3 different NSCLC patients harboring EGFR mutations.
   Highlighting profound alterations in cytokine secretion in PCs in
   comparison to cancer cells. B,C) KEGG and GO biological process
   analyses of the secretomics data in (A). The pathways related to
   cell‐cell communications and EGFR TKI resistance are highlighted in
   red. D) Representative images of proteome profiler cytokine XL array
   from the conditioned medium (CM) of PC and HCC827 cells are given. The
   expression of IL‐22, IL32, CCL5, IL17A, and CXC12 was up‐regulated in
   the CM of PCs as compared to that of HCC827 cells. E) Heatmap showing
   the RT‐PCR analysis of the relative expression of indicated cytokines
   in PCs as compared to HCC827 or PC9 cells. F) Western blot analysis was
   performed to evaluate IL32 expression in cancer cells and PCs derived
   from tumors of three EGFR‐mutated NSCLC patients. The quantification of
   the Western blot results is provided below the blots. The data are
   normalized to the control group. G) ELISA measurements were conducted
   to assess IL32 expression in the CM obtained from normal fibroblast
   (NF), human umbilical vein endothelial cell (EC), cancer cell lines
   (HCC827/PC9/NCI‐H1975), and pericyte (PC). H–J) Single‐cell RNA
   sequencing analysis of the published dataset ([87]GSE171145) derived
   from NSCLC patients harboring EGFR mutations. UMAP visualization map of
   all cell type clusters in tumors derived from EGFR mutant NSCLC
   patients. Each color represents the annotation phenotype of a single
   cluster (H). The circle plot shows the communication strength between
   interacting cells. The thicker the line represented, the higher the
   number of interactions (I). KEGG pathway enrichment analysis in
   IL32+Pericytes (J). K) The bar chart illustrates the percentage of
   pericyte‐IL32 positive blood vessels for each patient. L)
   Representative triple immunostaining images of IL32, α‐SMA, and CD34 on
   tumor sections from NSCLC patients with EGFR mutations, showcasing low
   or high pericyte‐IL32 expression, are provided. M) A high percentage of
   pericyte‐IL32 positive blood vessels correlated with poor
   progression‐free survival in EGFR‐mutated NSCLC patients (n = 20
   patients, our cohort). N) ELISA IL32 assays of serum samples from
   EGFR‐mutated cancer patients responsive or non‐responsive to
   third‐generation TKI treatment (n = 20 patients, our cohort). Violin
   plots represent mean ± S.E.M are given. NS: non‐significant difference.
   *p < 0.05, ***p < 0.001. (G) One‐way ANOVA. (M) Log‐rank (Mantel‐Cox)
   test. (N) Student's t‐test. Scale bars in (L) represent 100 µm.

   To assess the clinical significance of our findings, we conducted a
   single‐cell RNA sequencing analysis of publicly available data (GEO:
   [88]GSE171145).^[ [89]^23 ^] This analysis identified a subset of
   pericytes within EGFR‐mutated NSCLC patient‐derived tumors that
   exhibited high IL32 expression (Figure [90]1H; Figures [91]S1G,H,
   Supporting Information), predominantly in pericytes compared to other
   stromal cells like endothelial cells and fibroblasts (Figure [92]S1I,
   Supporting Information). Further analysis of cell‐cell communications
   via ligand‐receptor interactions unveiled robust communication between
   the subset of IL32‐positive pericytes and tumor cells expressing
   β5‐integrin (gene name: ITGB5) (Figure [93]1I; Figure [94]S1J,
   Supporting information). Moreover, this pericyte subset was
   significantly enriched in pathways related to the PI3K‐Akt signaling
   pathway, focal adhesion, cell adhesion molecules, and EGFR tyrosine
   kinase inhibitor resistance (Figure [95]1J; Figure [96]S1K, Supporting
   Information). Importantly, β5‐integrin is recognized as the putative
   binding receptor for IL32.^[ [97]^12 , [98]^20 ^] In contrast, the
   IL32‐negative pericyte subset exhibited a weaker association with
   cancer cells compared to the IL32‐positive pericyte subset and did not
   show enrichment in pathways observed in the IL32‐positive subset,
   particularly those related to EGFR tyrosine kinase inhibitor
   resistance. However, the IL32‐negative subset was enriched in pathways
   associated with metabolic processes (Figures [99]S1L,M, Supporting
   Information). To further corroborate our observations, we collected a
   sample cohort of advanced NSCLC patients with EGFR mutations who
   received third‐generation TKI treatment, with clinicopathological data
   (Table [100]S1, Supporting Information). We performed triple
   immunostaining for α‐SMA, CD34, and IL32 on tumor sections to quantify
   the percentage of pericyte‐IL32 positive blood vessels in each patient,
   indicating that IL32 was expressed in PCs (Figures [101]1K,L).
   Additional immunostaining experiments revealed the expression of
   β5‐integrin in cancer cells located around blood vessels within
   EGFR‐mutated tumors (Figure [102]S1N, Supporting Information).
   Consistently, previous studies showed that β5‐integrin is commonly
   expressed in various tumors, including NSCLC.^[ [103]^24 ^] Our data
   also indicated that a high percentage of pericyte‐IL32 positive blood
   vessels correlated with poor progression‐free survival in our patient
   cohort (Figure [104]1M). Notably, ELISA assays also showed that serum
   IL32 levels were higher in non‐responsive EGFR‐mutated patients treated
   with third‐generation TKIs compared to those who responded
   (Figure [105]1N). To corroborate this finding, we utilized the KM
   plotter database tool to analyze the correlation between the ratio of
   IL32 expression to pericyte marker expression and overall patient
   survival. Our results indicated that a high ratio of IL32 expression to
   pericyte marker expression, such as MCAM, PDGFRB, or ACTA2, was also
   linked to poor overall survival in NSCLC patients (Figure [106]S1O,
   Supporting Information). Interestingly, DNA sequencing results revealed
   that PCs isolated from EGFR‐mutated tumors did not themselves harbor
   the EGFR mutation (Figure [107]S1P, Supporting Information). This
   finding aligns with previous reports indicating that stromal cells have
   a lower mutation burden, and a more stable genome compared to cancer
   cells.^[ [108]^25 ^] In summary, our findings demonstrate that
   pericyte‐IL32 expression may determine the patient prognosis in
   EGFR‐mutated NSCLC patients and warrant further functional studies.

2.2. Pericytes Reduce TKI Sensitivity in EGFR‐Mutated Cancer Cells through
Paracrine Effect

   Our next objective was to investigate the potential of tumor‐derived
   pericytes (PCs) in modulating the sensitivity of NSCLC cell lines to
   TKI treatment. We conducted IC50 experiments with two distinct
   third‐generation TKI drugs, Almonertinib and Osimertinib, to
   demonstrate that exposing EGFR mutated cancer cell lines, such as
   HCC827, PC9 and NCI‐H1975 (T790M+ cell line) to conditioned medium (CM)
   from PCs resulted in reduced sensitivity to TKI treatment compared to
   cells exposed to CM from control cancer cells (Figures [109]2A,B;
   Figures [110]S2A,B, Supporting Information). Interestingly, our results
   demonstrated that when EGFR‐mutated cancer cells were exposed to CM
   from PCs, the effect on the sensitivity of first‐/second‐generation TKI
   drugs was not as pronounced as the effect on the third‐generation TKI
   drugs (Figure [111]S2C, Supporting Information). Furthermore,
   co‐culturing these NSCLC cell lines, labeled with GFP fluorescence,
   with PCs labeled with RFP fluorescence, also led to decreased
   sensitivity to TKI drugs compared to the control group, whereas the TKI
   treatment did not affect PC proliferation (Figures [112]2C,D; Figure
   [113]S2D, Supporting Information). Further CCK8 cell proliferation and
   apoptosis assays indicated that first‐ to third‐generation TKI drugs
   did not affect PC growth or apoptosis (Figures [114]S2E,F, Supporting
   Information). Consistently, RT‐PCR analysis showed that the expression
   of EGFR was lower in PCs as compared to cancer cells (Figure [115]S2G).
   We then investigated whether exposing these NSCLC cancer cell lines to
   CM from PCs influenced their colony formation and invasion ability in
   the presence of TKI drugs. Remarkably, our results revealed that
   exposure of cancer cells to CM from PCs enhanced their colony formation
   and invasion abilities in the presence of 3^rd generation TKI drugs
   compared to the group treated with CM from control cells
   (Figures [116]2E–H; Figures [117]S2H–J, Supporting Information).
   Conversely, when cancer cells were exposed to CM from PCs in the
   absence of TKI drugs, no significant impact on their proliferation,
   colony formation, or invasion ability was observed, as compared to
   cells exposed to CM from cancer cells (Figures [118]2C–H; Figures
   [119]S2D,H–J, Supporting Information). Additionally, our results showed
   that exposure to heat‐inactivated CM from PCs could no longer prohibit
   the inhibitory effect of TKI drugs on the colony formation and invasion
   ability of HCC827, PC9, and NCI‐H1975 cancer cell lines
   (Figures [120]2I,J; Figures [121]S2K–M, Supporting Information). This
   suggests that PCs may regulate TKI sensitivity in EGFR‐mutated NSCLC
   cell lines, possibly through paracrine signaling.

Figure 2.

   Figure 2
   [122]Open in a new tab

   Pericytes modulate the sensitivity of EGFR‐mutated cancer cells to TKIs
   through a paracrine mechanism. A,B) Osimertinib/Almonertinib IC50
   experiments of HCC827 cells treated with conditioned medium (CM) from
   HCC827 cells or PCs. C,D) Representative images of GFP fluorescently
   labeled HCC827 cells co‐cultured with/without RFP fluorescently tagged
   PCs in the presence or absence of Osimertinib/Almonertinib. Bar charts
   show the relative number of GFP‐positive cancer cells (top) or
   RFP‐positive PCs (bottom) in each experimental group. E,F)
   Representative crystal violet‐stained images of colony formation in
   each group are given. Bar charts show the relative colony number in
   each group. G,H) Transwell invasion assays of HCC827 cells treated with
   conditioned medium (CM) from HCC827 cells (HCC827‐CM) or PCs (PC‐CM) in
   the presence or absence of Osimertinib/Almonertinib. Bar charts show
   the relative number of invaded cells in each group. I) Representative
   crystal violet‐stained images of colony formation in each group are
   given. Bar charts show the relative colony number in each group. J)
   Representative images of crystal violet‐stained invaded cells of each
   group are given. Bar charts show the relative number of invaded cells
   in each group. K–O) Nude mice were subcutaneously injected with either
   HCC827 cells alone or co‐injected with HCC827 cells and pericytes (PCs)
   derived from tumors of patients with EGFR‐mutated NSCLC. Tumor‐bearing
   mice were then treated with a placebo or Osimertinib. Representative
   gross tumor image from each group is given (K). The line graph
   represents the tumor growth in each group over time (L). The bar chart
   shows the final tumor weight (M) or volume (N) in each group (n = 5
   mice per group). Representative images of CD34, α‐SMA, and Ki67
   triple‐immunostaining on tumor sections derived from each group are
   given. The violin plot shows the mean percentage of Ki67‐positive tumor
   cells near blood vessels in each group (O). NS: non‐significant
   difference. ^** p < 0.01, ***p < 0.001. C–J,M,N) One‐way ANOVA. C,D
   (bottom)) Students’ t‐test. L) Two‐way ANOVA. Scale bars in C,D)
   represent 200 µm. G,H,J) 100 µm. K) 1 cm. O) 10 µm.

   To validate our findings in an in vivo setting, we co‐injected HCC827
   cells with or without PCs into nude mice subcutaneously. The
   tumor‐bearing mice received treatment with either placebo or
   Osimertinib once a week for up to four weeks. Strikingly, our
   experiments demonstrated that Osimertinib treatment strongly suppressed
   tumor growth in mice injected with HCC827 cells alone. However, the
   repressive effect on tumor growth was notably diminished in mice
   co‐injected with HCC827 cells and PCs (Figures [123]2K–N).
   Interestingly, analysis of triple immunostaining experiments showed a
   higher number of Ki67‐positive tumor cells, particularly around
   vascular pericytes, in Osimertinib‐treated HCC827 and PC co‐injected
   tumors compared to Osimertinib‐treated tumors derived solely from the
   injection of HCC827 cells (Figure [124]2O). To examine the effect of
   PCs on cancer cell metastasis in vivo, we co‐injected luciferase‐tagged
   HCC827 cells with or without PCs into nude mice via the tail vein.
   Tumor‐bearing mice were treated with either placebo or Osimertinib,
   followed by bioluminescent imaging for analysis (Figures [125]S2N,O,
   Supporting Information). The results showed that co‐injection of cancer
   cells with PCs enhanced experimental lung metastasis as compared to the
   injection of cancer cells alone (Figures [126]S2N,O, Supporting
   Information). In summary, our findings suggest that PCs reduce the
   sensitivity of EGFR‐mutated cancer cells to TKI treatment through
   paracrine signaling.

2.3. Depleting IL32 Expression in PC Rescues its Paracrine Effect on EGFR
Mutant Cell Sensitivity to TKI Treatment

   To determine whether PCs affected EGFR mutant cell sensitivity to TKI
   through IL32, we knocked down IL32 expression in PCs using two
   different IL32 targeting siRNAs, and its silencing was confirmed by
   both western blotting and RT‐PCR (Figure [127]3A). The transfected
   cells were then used to perform a range of in vitro tumorigenicity
   assays. Drug sensitivity experiments showed that exposure of HCC827/PC9
   cells with CM from IL32‐depleted PCs reduced their TKI IC50 values when
   compared to the cells exposed with CM from non‐silencing siRNA (siNSC)
   transfected PCs (Figure [128]3B; Figure [129]S3A, Supporting
   Information). IC50 experiments also demonstrated that the use of an
   IL32‐neutralizing antibody reversed the repressive effect of CM from
   PCs on cancer cell sensitivity to third‐generation TKI drugs (Figure
   [130]S3B, Supporting Information). This observation was further
   confirmed by the co‐culturing experiment, indicating that co‐culturing
   EGFR mutant cells with IL32‐depleted PCs reduced the enhanced
   proliferation observed in the cells after co‐culture with siNSC
   transfected PCs, whereas there was no difference in the cell number and
   apoptosis between PCs transfected with either IL32 targeting siRNA‐1/‐2
   or siNSC (Figure [131]3C; Figures [132]S3C,D, Supporting Information).
   Furthermore, our results showed that exposure of cancer cells with CM
   from IL32‐depleted PCs reduced their colony formation and invasion
   ability in the presence of TKI treatment as compared to the cells
   exposed to CM from scramble transfected cells, whereas it had no effect
   in the absence of TKI treatment (Figures [133]3D–G; Figures [134]S3E–H,
   Supporting Information). Previous studies indicated that IL32 could
   bind integrin via its RGD motif to regulate breast cancer cell
   metastasis.^[ [135]^21 ^] Notably, β5‐integrin is known to contain the
   RGD binding domain.^[ [136]^12 , [137]^20 ^] We then investigated
   whether targeting β5‐integrin in cancer cells could reverse the
   paracrine effects of PCs on TKI drug sensitivity. Our results indicate
   that treating HCC827/PC9 cells with CM from PCs, followed by
   administration of the β5‐integrin inhibitor Cilengitide, or β5‐integrin
   depletion via shRNA, led to reduced IC50 values for TKI drugs compared
   to placebo‐treated cells (Figures [138]3H–J; Figures [139]S3I–K,
   Supporting Information). Significantly, Cilengitide treatment inhibited
   the augmented colony formation and invasive potential seen in
   EGFR‐mutated cancer cells treated with placebo, which were exposed to
   CM from PCs along with TKI drugs (Figures [140]3K,L; Figures
   [141]S3L,M, Supporting Information). In conclusion, our findings
   suggest that PCs influence the response of EGFR‐mutated cells to TKI
   treatment, potentially via the IL32‐β5‐integrin signaling pathway. This
   warrants deeper exploration and investigation.

Figure 3.

   Figure 3
   [142]Open in a new tab

   Disrupting IL32‐β5‐integrin paracrine signaling prevents
   pericyte‐mediated TKI tolerance in EGFR mutated cancer cells. A)
   Western blot and RT‐PCR analysis of the expression of IL32 in PCs after
   transfected with IL32 targeting siRNA‐1/‐2 or non‐silencing siRNA
   (siNSC). The quantification of the Western blot results is provided
   below the blots. The data are normalized to the control group. B)
   Almonertinib/Osimertinib IC50 experiments of HCC827 cells treated with
   conditioned medium (CM) from PCs transfected with either siNSC or IL32
   targeting siRNA‐1/‐2. C) Representative fluorescent images of GFP
   fluorescently labeled HCC827 cells co‐cultured with RFP fluorescently
   labeled siNSC or siIL32‐1/‐2 transfected PCs in the presence of
   Almonertinib/Osimertinib. Bar charts show the relative number of GFP or
   RFP‐positive cells in each group. D,E) Colony formation assays of
   HCC827 cells after being treated with CM from PCs transfected with
   either siNSC or IL32 targeting siRNA‐1/‐2 in the presence/absence of
   Almonertinib/Osimertinib. Bar charts show the relative colony number in
   each group. F,G) Transwell invasion assays of HCC827 cells after being
   treated with CM from PCs transfected with either non‐silencing control
   (siNSC) (PC‐siNSC‐CM) or IL32 targeting siRNA‐1/‐2 (PC‐siIL32‐1/‐2‐CM)
   in the presence/absence of Almonertinib/Osimertinib. Bar charts show
   the relative number of invaded cells in each group. H)
   Almonertinib/Osimertinib IC50 experiments of HCC827 cells after being
   treated with CM from PCs together with or without Cilengitide. I)
   Western blot and RT‐PCR analysis of the expression of β5‐integrin in
   each group. J) Almonertinib/Osimertinib IC50 experiments of HCC827
   cells stably transfected with scramble (HCC827‐scr) or β5‐integrin
   targeting shRNA (HCC827‐shβ5‐integrin) treated with CM from PCs. K)
   Colony formation assays of HCC827 cells after being treated with CM
   from HCC827 or PCs in the presence of Almonertinib/Osimertinib together
   with or without Cilengitide. Representative images of crystal
   violet‐stained colonies in each group are given. Bar charts show the
   relative colony number in each group. L) Transwell invasion assays of
   HCC827 cells after being treated with CM from HCC827 or PCs in the
   presence of Almonertinib/Osimertinib together with or without
   Cilengitide. Bar charts represent the relative number of invaded cells
   in each group. NS: non‐significant difference. ^** p < 0.01,
   ***p < 0.001. A,C–G,K,L) One‐way ANOVA. I) Student's t‐test. Scale bars
   in (C) represent 200 µm. F,G,L) 100 µm.

2.4. Pericyte‐Secreted IL32 Activates the β5‐Integrin‐Src‐Akt Cell Survival
Pathway, Impacting TKI Sensitivity in Cancer Cells

   To further dissect the underlying molecule mechanism of pericyte‐IL32
   mediated TKI sensitivity in cancer cells, we performed comparative
   proteomics analysis of HCC827 or PC9 cells after being treated with
   conditioned medium (CM) from either HCC827/PC9 cells or PCs in the
   presence of TKI treatment. Volcano plot analysis, as well as KEGG and
   GO biological process analyses of the proteomics data, revealed notable
   changes and enrichments in protein expression, pathways, and biological
   processes associated with the PI3K‐Akt signaling pathway, MAPK
   signaling pathway, EGFR tyrosine kinase inhibitor resistance, response
   to drug, integrin‐mediated signaling pathways, and cell‐cell adhesion
   mediated by integrins in cancer cells treated with PC‐conditioned
   medium alongside the TKI drug, as compared to the control group
   (Figures [143]4A–C; Figures [144]S4A–C, Supporting Information). These
   results suggested that pericyte‐IL32 may affect cancer cell sensitivity
   to TKI via its receptor integrin‐mediated signaling pathway. Indeed,
   previous studies have shown that β5‐integrin can regulate cancer cell
   survival and drug sensitivity through the Src‐Akt pathway.^[ [145]^26
   ^] We then conducted Western blot analysis on the downstream effectors
   of β5‐integrin, Src, and Akt, in cancer cells treated with conditioned
   medium (CM) from PCs or cancer cells, with or without TKI treatment.
   Our results show that exposing cancer cells to CM from PCs protected
   them from the inhibitory effects of TKI drugs on Src and Akt
   phosphorylation and activation, compared to cells exposed to CM from
   cancer cells (Figure [146]4D; Figures [147]S4D,E, Supporting
   Information). Consistently, previous studies indicated that TKI
   treatment could potentially inhibit EGFR‐mediated Src activation and
   its downstream Akt‐dependent cell survival/proliferation pathways in
   EGFR‐mutated cancer cells to reduce their survival, while the
   expression and activity of Src has been linked with TKI sensitivity.^[
   [148]^27 ^] In contrast, we demonstrated that exposing cancer cells to
   CM from PCs did not affect ERK phosphorylation, a key component of the
   MAPK signaling pathway, regardless of TKI treatment, compared to cells
   exposed to CM from cancer cells (Figures [149]S4F,G, Supporting
   Information). To determine whether PCs regulated Src‐Akt signaling in
   cancer cells via IL32‐β5‐integrin paracrine signaling, we treated
   cancer cells with CM from PCs transfected IL32 targeting siRNA or siNSC
   in the presence/absence of TKI drugs. Our results indicated that
   exposure of cancer cells with CM from siIL32 transfected PCs could no
   longer repress the inhibitory effect of TKI drugs on Src and Akt
   phosphorylation as compared to the cells exposed with CM from siNSC
   transfected PCs (Figure [150]4E; Figures [151]S4H,I, Supporting
   Information). Additionally, we showed that treatment with β5‐integrin
   inhibitor rescued the repressive effect of TKI drugs on Src and Akt
   phosphorylation in PC‐CM treated cancer cells as compared to
   placebo‐treated control group (Figures [152]4F,G; Figures [153]S4J,K,
   Supporting Information). Furthermore, we demonstrated that suppressing
   β5‐integrin expression in cancer cells decreased Src and Akt
   phosphorylation following treatment with CM from PCs in the presence of
   TKIs (Figure [154]4H; Figure [155]S4L, Supporting Information),
   suggesting the specific role of β5‐integrin in mediating cancer cell
   sensitivity to TKIs. In our clinical observations, immunohistochemistry
   results revealed a positive correlation between tumor‐p‐Src or p‐Akt
   and pericyte‐IL32 levels in our EGFR‐mutated patient cohort
   (Figures [156]4I,J). To further confirm the effector role of Src in
   PC‐IL32 mediated TKI sensitivity in cancer cells, we performed IC50 TKI
   experiments using EGFR‐mutated cancer cells treated with CM from PCs in
   the presence of TKI with or without Src inhibitor Dasatinib. The
   results showed that treatment with Dasatinib rescued the sensitivity of
   cancer cells to TKI treatment in the presence of CM from PCs (Figure
   [157]S4M, Supporting Information). In summary, our findings indicate
   that pericyte‐derived IL32 activates the β5‐integrin mediated Src‐Akt
   signaling pathway, thereby negating the inhibitory effects of TKIs on
   this pathway in EGFR‐mutated cancer cells.

Figure 4.

   Figure 4
   [158]Open in a new tab

   Pericyte‐secreted IL32 regulates β5‐integrin‐Src‐Akt pathway in EGFR
   mutated cancer cells, influencing TKI sensitivity. A) The volcano plot
   depicts the proteomics analysis of HCC827 cells following treatment
   with CM from HCC827 cells or PCs in the presence of Osimertinib. B,C)
   KEGG and GO pathway analyses of the proteomics data in (A). Pathways
   related to cell signaling, cell‐cell communications, and drug response
   are highlighted in red. D) Western blot analysis of the indicated
   proteins in HCC827 cells after being treated with CM from HCC827 or PCs
   in the presence or absence of Almonertinib/Osimertinib. The
   quantification of the Western blot results is provided below the blots.
   E) Western blot analysis of the indicated proteins in HCC827 cells
   after exposure with CM from PCs transfected with siNSC or siIL32 in the
   presence or absence of Almonertinib/Osimertinib. F,G) Western blot
   analysis was conducted to assess the indicated proteins in HCC827 cells
   following treatment with or without Almonertinib/Osimertinib, in
   combination with or without the β5‐integrin inhibitor Cilengitide. H)
   Western blot analysis was performed on the indicated proteins in HCC827
   cells transfected with β5‐integrin‐targeting shRNA (shβ5‐integrin) or
   scramble shRNA (scr) after being treated with CM from PCs in the
   presence of Almonertinib/Osimertinib. I) Representative
   immunohistochemical staining images of p‐Src and p‐Akt in tumor
   sections from each group are shown. Violin plot shows the staining
   intensity of p‐Src/p‐Akt in each group (n = 20 patients, our cohort).
   J) Correlation study between tumor‐p‐Src/p‐Akt and pericyte‐IL32 levels
   in our EGFR‐mutated NSCLC patient cohort (n = 20 patients, our cohort).
   I) Student's t‐test. J) Pearson correlation coefficient. Scale bars in
   (I) represent 100 µm.

2.5. YY1 Transcriptionally Regulates IL32 Expression and Secretion in PCs

   We next explored the molecular mechanism of IL32 up‐regulation in PCs.
   By performing a transcriptional binding site prediction study with the
   IL32 promoter, the results indicated that it contains 2 putative YY1
   (Yin‐Yang 1), one putative STAT3, and one putative ETS1 transcriptional
   factor binding site in the IL32 promoter (Figure [159]5A; Figure
   [160]S5A, Supporting Information). Interestingly, YY1, STAT3, and ETS1
   are transcriptional factors that have been linked with cancer
   progression and/or inflammatory responses.^[ [161]^28 ^] However, their
   role in regulating pericyte‐IL32 expression and secretion is unclear.
   We therefore silenced the expression of YY1, STAT3, or ETS1 in PCs by
   using siRNAs and performed western blot and/or RT‐PCR analysis of the
   expression of YY1/STAT3/ETS1 and IL32, indicating that the expression
   of YY1 and IL32 at the protein and mRNA levels was reduced in YY1
   depleted PCs as compared to siNSC transfected PCs (Figures [162]5B,C).
   In contrast, RT‐PCR analysis showed that silencing STAT3 or ETS1 did
   not affect IL32 expression in PCs as compared to siNSC transfected PCs
   (Figures [163]S5B,C, Supporting Information). To further determine the
   regulatory role of YY1 on IL32 promoter, we performed a chromatin
   immunoprecipitation experiment using YY1 antibody against siYY1 or
   siNSC transfected PCs, which were then subjected to RT‐PCR analysis.
   The results showed increased binding of YY1 to IL32 promoter through
   its predicted putative binding sites (sites 1 and 2) in siNSC
   transfected PCs as compared to siYY1 transfected PCs (Figure [164]5D).
   Moreover, a luciferase assay revealed that the promoter activity of
   IL32 in PCs was significantly enhanced in a dose‐dependent manner when
   co‐transfected with a luciferase reporter vector containing the
   full‐length IL32 promoter and a YY1 overexpression vector
   (Figure [165]5E). Importantly, our results showed that mutating the YY1
   putative binding site 1 or 2 significantly reduced the promoter
   activity of IL32 in PCs (Figure [166]5F). These results implicated that
   YY1 transcriptionally regulated IL32 expression in PCs.

Figure 5.

   Figure 5
   [167]Open in a new tab

   YY1 transcriptionally up‐regulates IL32 expression in pericytes. A)
   Promoter prediction study of IL32 reveals the presence of
   transcriptional factor YY1 putative binding sites. Schematic diagram
   showing the position of the two predicted YY1 putative binding sites.
   B,C) Western blot or RT‐PCR analysis of YY1 and IL32 expression in PCs
   transfected with siNSC or siYY1‐1/‐2. The quantification of the Western
   blot results is provided below the blots. D) Chromatin
   immunoprecipitation (ChIP) assay showed an increased binding of YY1 to
   both of its putative binding sites in siNSC transfected PCs as compared
   to siYY1 transfected PCs. E) The bar chart illustrates the relative
   luciferase activity in PCs following co‐transfection with a luciferase
   vector containing the full‐length IL32 promoter sequence and varying
   levels of YY1 overexpression plasmid. F) The bar chart displays the
   relative luciferase activity in PCs following co‐transfection with a
   luciferase vector containing either the full‐length IL32 promoter
   sequence (WT) or a promoter sequence with mutated proximal or distal
   YY1 putative binding sites, along with or without a YY1 overexpression
   plasmid. G) The bar chart shows the percentage of pericyte‐YY1 positive
   blood vessels in each patient. H) Representative triple immunostaining
   images of YY1, α‐SMA, and CD34 on tumor sections derived from each
   group are given. I) Correlation study between pericyte‐YY1 and
   pericyte‐IL32 expression in our NSCLC patient cohort (n = 20 patients).
   J) Almonertinib/Osimertinib IC50 experiments of HCC827 cells treated
   with CM from PCs transfected with either siNSC or YY1 targeting
   siRNA‐1/‐2. K) Representative fluorescent images of GFP fluorescently
   labeled HCC827 cells co‐cultured with RFP fluorescently labeled siNSC
   or siYY1‐1/‐2 transfected PCs in the presence of
   Almonertinib/Osimertinib. The bar chart shows the relative number of
   GFP (left) or RFP (right) positive cells in each group. L,M) Colony
   formation assays of HCC827 cells after being treated with CM derived
   from siNSC or siYY1‐1/‐2 transfected PCs in the presence/absence of
   Almonertinib/Osimertinib. The bar chart shows the relative colony
   number in each group. N,O) Transwell invasion assays of HCC827 cells
   after being treated with CM derived from siNSC or siYY1‐1/‐2
   transfected PCs in the presence/absence of Almonertinib/Osimertinib.
   The bar chart shows the relative number of invaded cells in each group.
   NS: non‐significant difference. *p < 0.05, ^** p < 0.01, ***p < 0.001.
   C–F,K–O) One‐way ANOVA. I) Pearson correlation coefficient. Scale bars
   in H,N,O) represent 100 µm. K) 200 µm.

   To examine the clinical significance of our findings, we also performed
   triple‐immunostaining of CD34, α‐SMA, and YY1 on tumor sections derived
   from our EGFR mutated NSCLC patient cohort. Our findings suggested a
   positive correlation between the expression levels of YY1 and IL32 in
   pericytes within these patients (Figures [168]5G–I). To determine the
   regulatory role of YY1 in PC‐mediated cancer cell sensitivity to TKI
   drugs, we performed co‐culture experiments with siYY1/siNSC‐transfected
   PCs and cancer cells. Strikingly, our results showed that exposure of
   cancer cells with CM from siYY1 transfected PCs reduced their IC50 TKI
   values as compared to the cells exposed with CM from siNSC transfected
   PCs (Figure [169]5J; Figure [170]S5D, Supporting Information) while
   co‐culturing cancer cells with siYY1 transfected PCs increased their
   sensitivity to TKI drugs as compared to the group co‐cultured with
   siNSC transfected PCs (Figure [171]5K; Figure [172]S5E, Supporting
   Information). In contrast, knocking down YY1 expression in PCs did not
   affect their proliferation and apoptosis (Figure [173]5K; Figures
   [174]S5E,F, Supporting Information). Furthermore, it also reduced the
   colony formation and invasion of cancer cells in the presence of TKI
   drugs as compared to the cells treated with CM harvested from
   placebo‐treated PCs (Figures [175]5L–O; Figures [176]S5G–J, Supporting
   Information). To confirm that IL‐32 is the downstream effector of
   YY1‐mediated PC effects on EGFR‐mutated cancer cell sensitivity to TKI,
   we exposed these cells to CM from YY1‐depleted PCs in the presence of
   TKI drugs, with or without IL32 treatment. The IC50 results indicated
   that the addition of IL32 increased the IC50 value of TKI drugs in
   EGFR‐mutated cancer cells treated with CM from YY1‐depleted PCs
   compared to the placebo‐treated group (Figures [177]S5K,L, Supporting
   Information). Our findings collectively suggest that YY1
   transcriptionally upregulates IL32 secretion from pericytes,
   influencing their paracrine effect on TKI sensitivity in cancer cells.
   This provides an initial explanation for the regulatory mechanism of
   IL32 expression within the tumor microenvironment.

2.6. Disrupting Pericyte‐Cancer Cell Crosstalk Rescues the Sensitivity of
EGFR Mutated Tumors to TKI, Regardless of Pericyte's Role in Supporting Blood
Vessels

   To further examine the regulatory role of pericytes in regulating tumor
   cell TKI sensitivity in vivo, 4–6 weeks old nude mice were
   subcutaneously injected with either HCC827 cells alone or co‐injected
   with HCC827 cells and PCs, which were then treated with placebo or TKI
   together with or without β5‐integrin inhibitor Cilengitide (Figure
   [178]6A). Remarkably, our findings revealed that blocking β5‐integrin
   with Cilengitide restored the response of tumors derived from
   co‐injection with HCC827 cells and PCs to TKI treatment, in contrast to
   the group treated with TKI only (Figures [179]6B–E). In contrast, the
   combined treatment with Cilengitide and TKI did not show any additional
   inhibitory effect on tumor growth in tumors derived from HCC827
   cell‐only injections (Figures [180]6B–E). Moreover, our
   immunohistochemistry demonstrated a reduction in p‐Src, p‐Akt, and
   Ki‐67 staining in HCC827 and PC co‐injected tumors treated with
   Cilengitide and TKI, compared to the group treated with TKI alone
   (Figures [181]6F,G). However, there was no difference in the staining
   intensity of these markers in HCC827‐alone injected tumors treated with
   Cilengitide and TKI compared to the TKI‐only group (Figures [182]6F,G).
   Given the pivotal role of pericytes in regulating blood vessel
   functions,^[ [183]^12b ^] we conducted in vivo ultrasound microbubble
   contrast imaging to investigate whether the co‐injection of PCs and
   tumor cells impacted blood vessel functions and influenced TKI drug
   sensitivity in vivo. Our results showed no significant difference in
   blood vessel perfusion and flow between the tumors derived from either
   co‐injection of PCs and tumor cells or injection of tumor cells alone
   even in the presence or absence of TKI treatment was observed
   (Figures [184]6H–J). To further confirm our observation, we also
   co‐injected HCC827 cells with PCs transfected with either scramble
   (scr) or IL32 targeting shRNA (shIL32) subcutaneously. Tumor‐bearing
   mice were then treated with or without TKI and their tumor growth was
   monitored over time (Figures [185]S6A–E, Supporting Information). The
   results indicated that depleting IL32 expression in PCs increased the
   sensitivity of tumors to TKI treatment as compared to tumors derived
   from co‐injection of HCC827 cells and scramble transfected PCs (Figures
   [186]S6A–E, Supporting Information). In contrast, there was no
   significant difference in tumor growth between
   HCC827/scramble‐transfected PC and HCC827/shIL32‐transfected PC
   co‐injected tumors in the absence of TKI treatment (Figures [187]S6A–E,
   Supporting Information). Overall, these results suggested that PCs
   conferred TKI sensitivity in EGFR‐mutated tumors through
   IL32‐β5‐integrin paracrine signaling regardless of their blood vessel
   supporting role.

Figure 6.

   Figure 6
   [188]Open in a new tab

   Inhibition of β5‐integrin restores sensitivity to TKI in co‐injected
   tumors of HCC827 cells and PCs. A) The schematic diagram illustrates
   the treatment protocol for our animal experiments. Nude mice were
   subcutaneously injected either with HCC827 cells alone or co‐injected
   with HCC827 cells and PCs. Tumor‐bearing mice were then treated with
   either a placebo or Osimertinib, with or without Cilengitide, followed
   by microbubble‐based ultrasound imaging and tumor harvesting. B)
   Representative gross tumor image from each group is given. C) The line
   graph represents the tumor growth in each treatment group. D,E) The bar
   chart shows the final tumor weight (top) or volume (bottom) in each
   group. F) Representative immunohistochemistry images of p‐Src, p‐Akt,
   and Ki67 staining in each experimental group are given. G) Bar charts
   show the relative staining intensity of p‐Src/p‐Akt/Ki67 in each group.
   H) Microbubble contrast ultrasound imaging of subcutaneous tumors
   derived from mice, which were treated with either placebo or
   Osimertinib. I,J) Bar charts show the relative blood flow or perfusion
   within the whole tumor or tumor core in each group. NS: non‐significant
   difference. ^** p < 0.01, ***p < 0.001. C) Two‐way ANOVA. D,E,G,I,J)
   One‐way ANOVA. Scale bars in B,H) represent 1 cm. F) 100 µm.

   In conclusion, we identify PCs regulate EGFR mutant lung cancer cells
   via a paracrine effect. Clinically, high expression of pericyte‐IL32 is
   associated with poor prognosis in NSCLC patients harboring EGFR
   mutations. Mechanistically, YY1 transcriptionally up‐regulates IL32
   secretion from PCs, which then activates the β5‐integrin‐Src‐Akt
   signaling pathway in EGFR‐mutant cancer cells to confer their TKI
   sensitivity. Our study identifies a promising therapeutic target to
   enhance TKI responsiveness in NSCLC, highlighting the need for further
   clinical investigation into its potential efficacy.

3. Discussion

   NSCLC, a leading cause of global cancer‐related deaths, has seen a
   shift from traditional chemotherapy to targeted therapies, particularly
   for EGFR mutations.^[ [189]^1 ^] Although first‐generation TKIs show
   initial efficacy,^[ [190]^3b ^] acquired resistance, often via the
   T790M mutation, poses challenges.^[ [191]^8b ^] Second‐generation TKIs
   initially promising against T790M in preclinical studies, encountered
   dose‐limiting toxicity in clinical trials.^[ [192]^7b ^] Recently,
   third‐generation EGFR inhibitors like Almonertinib and Osimertinib have
   emerged, targeting T790M mutations more effectively while preserving
   wild‐type EGFR function, gaining approval for NSCLC treatment.^[
   [193]^29 ^] Notably, third‐generation TKIs have also been approved as a
   first‐line treatment for patients with advanced NSCLC harboring EGFR
   mutations.^[ [194]^10 ^] Despite these advances, some patients remain
   poorly responsive, with unclear mechanisms,^[ [195]^11b ^] likely
   influenced by the stromal effect on cancer cell sensitivity to TKI
   drugs.

   Recent research, both from our team and other scientific groups, has
   shed light on the crucial role of stromal cells, such as pericytes
   (PCs), in influencing the growth of cancer cells and their response to
   chemotherapy through paracrine effects, which are independent of their
   traditional function in supporting blood vessels.^[ [196]^12 , [197]^16
   ^] However, the impact of PCs on the sensitivity of cancer cells to
   targeted therapies remains unclear. To investigate this, we conducted
   secretomics analysis and cytokine arrays using PC‐derived CM from NSCLC
   patients with EGFR mutations. We compared these results with those from
   cancer cells and found a specific increase in the proinflammatory
   factor IL32 in PCs. Although PCs are commonly recognized as cells that
   support blood vessels, they are also known to regulate inflammatory
   responses through their secretome.^[ [198]^12 , [199]^30 ^] In line
   with this, IL32 has been shown to influence cancer cell survival and
   apoptosis,^[ [200]^21 ^] although the underlying mechanism is unclear.
   Additionally, its role in modulating EGFR‐mutated cancer cell
   sensitivity to TKIs has not been explored previously. It was unclear
   whether PCs are the source of IL32 in the tumor microenvironment and
   whether this is correlated with cancer progression. Notably, lung
   cancer cells express the IL32 putative receptor, RGD binding receptor
   β5‐integrin, which plays a key role in regulating the Src‐Akt
   pathway,^[ [201]^26 , [202]^31 ^] a known downstream target of TKIs.^[
   [203]^27 , [204]^32 ^]

   Our clinical investigations revealed a strong link between
   pericyte‐IL32 expression and prognosis in EGFR‐mutated NSCLC patients
   undergoing third‐generation TKI treatment, suggesting it may indicate
   their response to treatment. This underscores the clinical relevance of
   our findings and highlights the potential impact of stromal cells on
   the efficacy of targeted therapy in NSCLC. Subsequent functional assays
   showed that PCs could regulate the sensitivity of EGFR‐mutated cancer
   cells to TKIs through paracrine effects. Depleting pericyte‐derived
   IL32 or blocking its putative receptor β5‐integrin could reverse their
   inhibitory effect on TKI sensitivity in cancer cells. Consistent with
   this finding, RGD binding integrin has previously been linked to the
   regulation of EGFR‐mutated cancer cell sensitivity to first‐generation
   TKIs.^[ [205]^22 ^] However, the mechanisms underlying its functional
   role in regulating cancer cell sensitivity to TKIs have remained
   unclear until now.

   To elucidate IL32 upregulation in PCs, we predicted transcriptional
   binding sites and performed chromatin immunoprecipitation with the IL32
   promoter. YY1 was identified as a regulator of IL32 expression in PCs,
   and YY1 silencing countered the inhibitory effect of PCs on cancer cell
   TKI sensitivity. Interestingly, YY1, a transcription factor, is highly
   expressed in various cancers and is implicated in cell survival and
   apoptosis.^[ [206]^33 ^] However, its role in TKI sensitivity has not
   been investigated until now. Further proteomics analysis revealed
   enrichment in integrin‐related pathways in cancer cells treated with
   PC‐conditioned medium and TKIs. PC‐conditioned medium countered TKI
   effects on downstream EGFR effectors Src and Akt phosphorylation,
   reversible by Src or β5‐integrin inhibition/depletion, while our animal
   works also reveal that the co‐injection of cancer cells with PCs
   diminishes the efficacy of TKIs in limiting tumor burden, irrespective
   of their blood vessel functions, probably by repressing the inhibitory
   effect of TKIs on EGFR‐Src‐Akt mediated cell survival/proliferation
   pathways. Previous studies consistently demonstrate that TKIs target
   the Src‐Akt signaling pathway, a critical downstream target essential
   for their impact on cancer cell survival.^[ [207]^27 , [208]^34 ^]
   Moreover, it is recognized that an unidentified signal triggering
   Src‐mediated Akt activation influences TKI sensitivity in cancer
   cells.^[ [209]^27c ^] Our findings demonstrate that PCs can circumvent
   the inhibitory effect of TKIs on Src‐Akt signaling through
   IL32‐β5‐integrin paracrine signaling. Interestingly, in comparison to
   first‐ and second‐generation TKIs, the paracrine effect of PCs on TKI
   drugs is more distinct in the case of third‐generation TKI drugs.
   Furthermore, the utilization of the clinically approved Src inhibitor
   Dasatinib can counteract the suppressive paracrine effect of PCs on the
   sensitivity of third‐generation TKIs in EGFR‐mutated cancer cells. This
   highlights the importance of focusing on the role of PCs in modulating
   the sensitivity of third‐generation TKIs. Additionally, we demonstrate
   that IL32 depletion or the use of integrin inhibitors can effectively
   counteract the inhibitory effect of PCs on cancer cell sensitivity to
   TKI in vivo. This suggests the potential for repurposing integrin/Src
   inhibitors as a strategy to enhance the efficacy of TKI‐targeted
   therapy.

4. Conclusion

   In conclusion, our study reveals the previously unrecognized role of
   pericytes in modulating the sensitivity of EGFR‐mutated cancer cells to
   TKI treatment. This discovery urges a reevaluation of our strategy for
   treating EGFR‐mutated NSCLC patients with elevated pericyte‐IL32
   expression. Further clinical investigations are warranted to explore
   the potential of combining β5‐integrin inhibitor and third‐generation
   TKIs for these patients.

5. Experimental Section

Ethics Approval and Patient Consent Statement

   The collection and usage of patient samples were carried out with the
   written informed consent of the patients and received approval from the
   ethical review committee at Sun Yat‐sen Memorial Hospital
   (SYSKY‐2023‐903‐01). All clinical samples were anonymized in compliance
   with local ethical guidelines, as mandated by the Declaration of
   Helsinki. Clinicopathological data are provided in Table [210]S1
   (Supporting information).

Clinical Sample and Data Analysis

   All tissue and blood samples from NSCLC patients with EGFR mutations,
   treated with third‐generation TKIs, were obtained from Sun Yat‐sen
   Memorial Hospital and confirmed by pathological examination. Pericytes
   (PCs) were isolated from tumor tissues of patients harboring EGFR
   mutations, including delE746_A750, L858R, or T790M mutations.
   Immunostaining followed established protocols.^[ [211]^12b ^] Briefly,
   paraffin‐embedded tissues were sectioned, de‐waxed, and subjected to
   antigen retrieval in boiling 10 mm citrate buffer pH 6.0. Sections were
   washed in PBS and blocked in 1% normal goat serum (NGS) with 0.1%
   Triton X‐100 for 1 h. Subsequently, sections were triple‐immunostained
   using anti‐alpha‐smooth muscle actin Cy3‐conjugated
   (Sigma–Aldrich‐#C6198), anti‐CD34 (Biolegend‐#343610), and either
   anti‐IL32 (Abcam‐#ab37158), anti‐Integrin β5 (Cell Signaling
   Technology‐#3629S) or anti‐YY1 antibodies (Cell Signaling
   Technology‐#46395). After three PBS washes, sections were incubated
   with secondary antibodies diluted 1:200 in 1% NGS 0.1% TX‐100, washed
   thrice, and mounted with Prolong Gold Anti‐Fade Mountant with DAPI
   (ThermoFisher Scientific‐#[212]P36931). For data analysis, the
   percentage of IL32 or YY1 and alpha‐smooth muscle actin
   (α‐SMA)‐double‐positive blood vessels was calculated by dividing the
   number of IL32 or YY1 and α‐SMA‐double‐positive blood vessels by the
   total number of α‐SMA‐positive blood vessels. Patient data were
   expressed as those with either lower or higher than the mean percentage
   of blood vessels positive for pericyte‐derived IL32/YY1 (IL32:44%;
   YY1:48%).

   For KM plotter database analysis, the ratio of IL32 to pericyte marker
   expression in NSCLC patients was calculated following methods described
   previously.^[ [213]^12b ^] In brief, this calculation entailed using
   selected gene probes in the ‘use multiple genes > use ratio of two
   genes’ option within the KM plotter database tool, which was carried
   out according to the instructions of the web tool developer.^[ [214]^35
   ^]

Single‐Cell RNA Sequencing Analysis

   Clinical information and data was retrieved from the GEO database
   ([215]GSE171145),^[ [216]^23 ^] specifically selecting samples with
   EGFR mutations in NSCLC for the analysis. The scRNA‐seq data underwent
   analysis using the R package “Seurat” (version 4.3.0) in R (version
   4.3) as described previously.^[ [217]^23 ^] After quality control
   filtering out low‐quality cells and performing batch effect correction,
   22,264 cells and 36,128 genes were left for further analysis. Then,
   highly variable genes (HVGs) were selected for principal components
   analysis (PCA), followed by Uniform Manifold Approximation and
   Projection (UMAP) dimension reduction. Clusters were identified using
   the FindNeighbors and FindClusters function, the clusters were
   annotated based on the average expression of curated gene marker sets
   representing major cell types. The R package “copykat” (version 1.1.0)
   was utilized to analyze copy number variation (CNV) in the epithelium
   and identify malignant cells. For cell–cell communication analysis, the
   R package “CellChat” (version 1.5.0) was employed to explore
   intercellular interactions, particularly between pericytes and other
   cell types in the tumor microenvironment.

Drugs

   EGFR‐TKI drugs, including Afatinib, Gefitinib, Osimertinib, and
   Almonertinib, as well as Cilengitide, were obtained from Selleck
   Chemicals, while Dasatinib was purchased from Sigma–Aldrich.

Cell Line and Tissue Culture

   HCC827 (RRID:CVCL_2063), PC9 (RRID:CVCL_B260) and NCI‐H1975
   (RRID:CVCL_1511) cancer cell lines carry either the EGFR delE746_A750
   mutation or the EGFR L858R/T790M double mutation, were procured from
   the American Type Culture Collection (ATCC) and cultured in RPMI1640
   medium supplemented with 10% FBS and 1% penicillin and streptomycin and
   maintained at 37 °C in a 5% CO[2] incubator. Human umbilical vein
   endothelial cells (HUVEC) were isolated and cultured in endothelial
   cell medium (ScienCell‐#1001) according to the manufacturer's
   instruction; GFP fluorescently labeled HCC827 and PC9 cells were
   generated by infecting the cells with GFP overexpressing lentiviruses
   and subsequently selecting with puromycin (2 µg ml^−1) until the cells
   expanded. Pericytes derived from tumor tissues of NSCLC patients
   with/without EGFR mutations were isolated and cultured following the
   methodology outlined in a previous publication,^[ [218]^12b ^] and they
   were maintained in commercially available pericyte medium
   (ScienCell‐#1201) and cultured at 37 °C in a 5% CO[2] incubator. DNA
   sequencing of these PCs to determine the presence of known EGFR
   mutations was conducted by Guangzhou IGE Biotechnology LTD. The PCs
   used in the experiments were at a passage number below 8. To obtain
   mCherry fluorescently labeled PCs, PCs were infected with mCherry
   overexpressing lentiviral constructs and subsequently selected using
   puromycin (2 µg ml^−1). For the heat inactivation experiment, it was
   done as described previously.^[ [219]^36 ^] Briefly, the conditioned
   media (CM) collected from PCs was subjected to heat treatment at 56 °C
   for 2 h. Subsequently, the CM was cooled down to 37 °C before being
   used. The pericyte culture underwent regular mycoplasma testing
   throughout this study.

Cell Viability and Apoptosis Assays

   The impact of TKI drugs on cancer cell survival or PC proliferation in
   vitro was assessed using the Cell Counting Kit‐8 assay kit (KeyGen
   BioTECH‐#KGA317). In brief, HCC827 cells (5000 cells/well), PC9 cells
   (3000 cells/well), or PCs (3000 cells/well) were seeded in a 96‐well
   plate containing RPMI 1640 medium supplemented with 10% FBS and 1%
   penicillin and streptomycin or pericyte medium. The following day,
   after the cancer cells had adhered to the well, they were exposed to
   various drug concentrations or control, as specified. For PCs, they
   were exposed to the IC50 value of each TKI drug, which was determined
   by using HCC827 cells. After 3 days of treatment or every 12 h within a
   period of up to 72 h, CCK8 reagent (10 µL/well) was added to the cells
   and incubated for 2 h before concluding the experiment. Subsequently,
   the plate was placed in a Tecan Spark® 10M Multimode plate reader to
   measure the OD value at 450 nm.

   For neutralization experiments, a neutralizing antibody against human
   IL32 (Rockland, USA, 500 ng ml^−1) or corresponding IgG antibody was
   pre‐incubated at 37 °C with the supernatant for 1 h before doing IC50
   assays. For the Src inhibition experiment, cancer cells were treated
   with or without 0.1 µm Dasatinib (Sigma–Aldrich‐#SML2589) in the
   presence of TKI at indicated doses. For the YY1 downstream effector
   experiment, cancer cells were exposed to CM from siYY1 transfected PCs
   in the presence of TKI drugs at indicated doses with or without
   25 ng ml^−1 of recombinant human IL‐32 (Biolegend‐#551002). Regarding
   the apoptosis assays, the proportions of viable cells and apoptotic
   cells within the entire cell population after indicated treatments were
   evaluated using a FITC Annexin V apoptosis detection kit with PI
   (BioLegend‐#6409914).

Transwell Invasion Experiment

   HCC827, PC9, and NCI‐H1975 cells were subjected to serum starvation for
   24 h. Subsequently, HCC827 (4.5 × 10^4 cells/well), PC9 (3.5 × 10^4
   cells/well) or NCI‐H1975 (4.5 × 10^4 cells/well) were collected and
   suspended in 500 µL of condition medium. These cells were then seeded
   into the upper compartments of transwell inserts with an 8 µm pore size
   (Corning) and coated with a layer of Matrixgel matrix
   (Corning‐#354230). Following this, 750 µL of RPMI1640 medium
   supplemented with 10% FBS was added to the lower compartments. After
   ensuring the absence of bubbles, the transwells were placed in a CO[2]
   incubator at 37 °C for 48 h. Subsequently, the chambers and the liquid
   in the lower compartments were discarded, and 1 mL of 4% PFA per well
   was added to fix the cells at room temperature for 15 min. The cells
   were then rinsed with PBS and incubated with 1 mL of 0.1% crystal
   violet dye at room temperature for 1 h. The chambers were removed using
   tweezers, and rinsed three times, and any cells that had not migrated
   beyond the chamber were gently wiped with a cotton swab. The chambers
   were then inverted to dry, and pictures were taken for subsequent
   analysis.

Colony Formation Experiment

   HCC827 cells (3000 cells/well) and PC9 cells (2000 cells/well) were
   initially seeded onto a 6‐well plate and allowed to adhere overnight.
   Subsequently, they were exposed to 2 mL of conditioned medium derived
   from HCC827 cells, PC9 cells, or PCs in the presence of vehicle or TKI
   drugs for 72 h. The cells underwent medium replacement with fresh
   medium containing either vehicle or TKI drugs twice a week and were
   continually cultured for up to 10 days. After this period, the cells
   were washed with PBS twice, fixed with 1 mL of 4% PFA at room
   temperature for 10–15 min, and treated with 0.1% crystal violet dye for
   1 h. Finally, they were washed with 1 mL of ddH[2]O three times, dried,
   and subjected to imaging under a light microscope to enumerate the
   number of colonies (with each colony comprising 50 cells) in each
   experimental condition.

siRNA/shRNA Transfection

   For siRNA transfection experiment, a total of 3.5×10^5 PCs were seeded
   in T25 flasks overnight and then transfected with 50 nM
   YY1/IL32/ETS1/STAT3 targeting siRNA molecules or non‐silencing control
   siRNA (siNSC) (purchased from IGE bio), using Lipofectamine RNAiMAX
   (ThermoFisher Scientific‐#13778150) as per the manufacturer's
   instructions. After 48 h of transfection, both the cells and the
   culture medium were collected for subsequent analysis. To conduct shRNA
   experiments targeting β5‐integrin/IL32, specific shRNA sequences
   designed for this purpose were acquired and integrated them into the
   psi‐LVRU6P vector obtained from GeneCopoeia. As a comparison, a
   scrambled shRNA sequence was also included in the psi‐LVRU6P vector.
   Following transfection of the recombinant constructs along with
   lentivirus packaging vectors into 293T cells using Lipofectamine 3000,
   viral supernatant was collected and stored it at −80 °C after
   filtration. The cancer cells or PCs were then infected with the viruses
   in the presence of polybrene and subjected to puromycin selection for
   two weeks. The resulting selected cells were expanded and utilized for
   subsequent experiments as needed. The sequences of siRNA or shRNA used
   in this study are given in Table [220]S2 (Supporting information).

RNA Extraction and RT‐PCR Analysis

   Cells were collected and washed with PBS before being transferred to
   1.5 mL EP tubes. After centrifugation, the supernatant was discarded,
   and the cells were resuspended in 1 mL TRIZOL
   (Biotopped‐#[221]TOP04079) and incubated on ice for 5 min. To each
   sample, 200 µL of chloroform was added, and the mixture was vigorously
   shaken for 15 s and allowed to stand at room temperature for 2 min. The
   samples were then centrifuged at 4 °C, 16000 g, for 15 min. The upper
   supernatant was carefully transferred to a new EP tube, and an equal
   volume of cold isopropyl alcohol was added. After gentle mixing, the
   samples were incubated on ice for 10 min. Following centrifugation at
   4 °C, 16000–18000 g, for 10 min, the supernatant was discarded, and the
   RNA pellet was washed and precipitated with 1 mL of cold 75% ethanol.
   After another centrifugation at 16000–18000 g for 5 min, the
   supernatant was removed, and the RNA pellet was air‐dried at room
   temperature. DEPC water was added to dissolve the samples, and RNA
   concentration was determined using a Nandrop spectrophotometer. For
   reverse transcription, 1 µg of RNA template was taken, and 5×
   All‐in‐one RT SuperMix, Enzyme Mix, and RNase‐free H[2]O (Vazyme‐#R333)
   were added according to the manufacturer's instructions. The mixture
   was then centrifuged and subjected to a PCR apparatus for the reaction.
   The cDNA obtained from reverse transcription was diluted, with
   generally 20 ng taken as the cDNA template. To this, a 2× SYBR Green
   mix (Yeasen‐#11201ES08) and the appropriate primers were added,
   followed by the addition of water. After thorough mixing, the samples
   were loaded onto the PCR plates and centrifuged. The qPCR analysis was
   performed using the Roche Light Cycler480 II machine. The sequences of
   primers used in this study are given in Table [222]S3 (Supporting
   information).

Western Blotting

   The supernatant from PCs or cancer cells in 10 cm culturing dishes or
   6‐well plates was discarded, and the cells were subsequently washed
   twice with cold PBS. Then, NP40 lysis buffer was added, supplemented
   with protease inhibitor (Merck Millipore‐#539131, 1:100), and
   phosphatase inhibitor cocktails (Merck Millipore‐#524625, 1:100). The
   cells were gently scraped using a cell scraper. Afterward, the lysate
   was centrifuged at a low temperature, and the supernatant was
   collected. Protein concentrations were determined according to the
   instructions of the BCA kit (Invitrogen‐#23225). For further analysis,
   an equal amount of protein from each sample was heat‐denatured and
   resolved on 10–12% SDS‐PAGE gels. These gels were then electroblotted
   onto polyvinylidene difluoride membranes (Merck Millipore‐#3010040001).
   The membrane was blocked in TBS with 0.1% Tween 20 (TBST) containing 5%
   milk for 1 h at room temperature. Subsequently, the membrane was probed
   with the following primary antibodies (diluted 1:1000) overnight at
   4 °C: anti‐IL32 (Abcam‐#ab37158), anti‐β5‐integrin (Cell Signaling
   Technology‐#3629S), anti‐p‐Src (Cell Signaling Technology‐#2101S),
   anti‐total Src (Cell Signaling Technology‐#2109S), anti‐total Akt (Cell
   Signaling Technology‐#4691S), anti‐p‐Akt (Cell Signaling
   Technology‐#4060S), anti‐YY1 antibodies (Cell Signaling
   Technology‐#46395), anti‐total ERK (Cell Signaling Technology‐#9102S),
   anti‐p‐ERK (Cell Signaling Technology‐#4370S), anti‐HSC70 (Santa
   Cruz‐#sc‐7298) and β‐actin (Santa Cruz‐#sc‐47778) (used as a loading
   control). After washing with TBST, the membrane was incubated with
   horseradish peroxidase‐conjugated anti‐rabbit secondary antibodies
   (Cell Signaling Technology‐#7074) for 1 h at room temperature, followed
   by three washes with TBST. The membranes were visualized using an
   enhanced chemiluminescence (ECL) system (Merck Millipore‐#WBKLS0500).
   Western blot quantification was performed using ImageJ software. The
   data were normalized to the control group.

Cytokine Array Experiment

   The Cytokine array experiment was done using the Proteome Profiler
   Human XL Cytokine Array Kit according to the manufacturer's instruction
   (Biotechne‐#ARY022B) and as described previously.^[ [223]^37 ^] The
   conditioned medium from cancer cells or PCs was harvested and used for
   the experiment.

ELISA Experiment

   The IL32 ELISA was performed according to the manufacturer's
   instructions (American Research Products Inc.‐#ELK2685). Briefly, 2×10⁴
   normal fibroblasts (NF), HUVECs, PCs, HCC827, or PC9 cells were seeded
   per well in 6‐well plates and incubated overnight. The cells were then
   cultured in a serum‐free medium, and the supernatant was collected
   after 48 h for ELISA analysis. Additionally, serum from EGFR‐mutated
   patients, both responsive and non‐responsive to third‐generation TKI
   drugs, was tested according to the manufacturer's instructions.

Animal Model

   All animal procedures in this study received approval from the
   institutional animal care and use committee of Sun Yat‐sen University
   (SYSU‐IACUC‐2023‐001796). In brief, 4–6 weeks old female nude mice
   (purchased from Guangdong medical laboratory animal center) were
   subcutaneously injected with 3×10^6 HCC827 cells alone or co‐injected
   with 3×10^6 HCC827 cells and 1×10^6 PCs in 100 µL mixture of PBS and
   Matrigel (BD Biosciences) (1:1 ratio). When the mean tumor size reached
   ≈200 mm^3, the tumor‐bearing mice were randomly assigned to different
   treatment groups, which included: 1. Vehicle alone; 2. Osimertinib
   alone (administered orally by intragastric injection at a dose of
   5 mg kg^−1 once a week); 3. Cilengitide alone (administered
   intraperitoneally at a dose of 5mg kg^−1 once a week); 4. Osimertinib
   plus Cilengitide. Tumor diameters were measured every 4–5 days using
   calipers, and tumor volumes were calculated using the formula
   L×W^2×0.52, where L represented the tumor length and W represented the
   tumor width. Subcutaneous tumors from each experimental group were also
   assessed via ultrasound before euthanizing the mice. For ultrasound
   imaging and blood vessel function analysis, super‐resolution ultrasound
   image acquisition was conducted in accordance with the Vevo 2100
   ultrasound machine manufacturer's instructions (Visual Sonics),
   employing the MS250 probe. Ultrasound‐enhanced contrast images were
   utilized to detect blood flow and tumor vascular perfusion. In brief,
   microbubble contrast reagents were injected into the tumor‐bearing mice
   through the tail vein using a tail vein microinjection catheter at
   100 µL per mouse. Data analysis was performed using the VevoCQ‐enhanced
   quantitative software. Throughout the experiments, the animals were
   kept in controlled conditions, which included maintaining an ambient
   temperature of 22–24 °C, controlling humidity levels between 40% and
   70%, and following a 12‐h light/dark cycle. To investigate the role of
   IL32 in PC‐mediated TKI sensitivity in cancer cells in vivo, nude mice
   were co‐injected subcutaneously with 3×10⁶ HCC827 cells and 1×10⁶ PCs
   transfected with either IL32‐targeting shRNA or scramble shRNA. The
   tumor‐bearing mice were then randomly assigned to placebo or
   Osimertinib treatment groups. For experimental metastasis studies, nude
   mice were injected via the tail vein with 0.7×10⁶ luciferase‐tagged
   HCC827 cells, with or without 0.3×10⁶ PCs. 10 days post‐injection, the
   tumor‐bearing mice were then randomly allocated to placebo or
   Osimertinib treatment groups, and tumor burden was assessed by
   bioluminescent imaging.

Immunohistochemistry for Mouse Tumor Tissues

   Mouse tumor tissues were fixed in formalin for 24 h and transferred to
   70% ethanol. The tissues were then paraffin‐embedded, sectioned,
   dewaxed, and unmasked in boiling 10 mm citrate buffer pH 6.0. 5 mm
   sections were washed three times in PBS, blocked in 1% normal goat
   serum (NGS) 0.1% Triton X‐100 (TX‐100) for 1 h. The tissue sections
   were immunostained with primary antibodies, including anti‐p‐Src (Cell
   Signaling Technology‐#2101S), anti‐p‐Akt (Cell Signaling
   Technology‐#4060S), and anti‐Ki67 (Servicebio
   Technology‐#[224]GB111499). Subsequently, sections were incubated with
   enzymatic Avidin–Biotin Complex (ABC) and visualized with
   3,3’‐diaminobenzidine (DAB) staining (Vector Laboratories), followed by
   counterstaining with hematoxylin. Each section was then assigned a
   score ranging from 0 to 3 based on both the area and intensity of
   staining. Specifically, the final staining score for each slide was
   calculated as the product of the area score and the intensity score.
   For Ki67 staining quantification, it was done as previously
   described.^[ [225]^38 ^] Briefly, Ki67‐positive tumor cells located
   within a perivascular distance of 50 µm from CD34/α‐SMA double‐positive
   blood vessels were quantified as a percentage of perivascular
   DAPI‐positive nuclei.

Secretomics and Proteomics Analysis

   The secretomics and proteomics analyses were conducted as previously
   described with some modifications.^[ [226]^39 ^] For secretomics
   analysis, HCC827 cells or PCs (2×10^6 cells/dish) were seeded in a
   10 cm culture dish and left undisturbed overnight. Subsequently, the
   cells were exposed to a serum‐free medium, and after 48 h, their
   supernatants were collected. These collected supernatants were then
   subjected to a vacuum apparatus for low‐temperature drying.
   Subsequently, the dried precipitate was re‐suspended using 50 mm
   NH[4]HCO[3] and quantified with a BCA protein quantification kit. From
   each sample, 50 µg of protein was taken, and four times its volume of
   acetone was added to precipitate the proteins. The mixture was gently
   mixed by inversion 6–8 times and then stored at either −30 °C overnight
   or −80 °C for 2 h. The samples were then removed and centrifuged at
   4 °C and 16,000 g for 30 min. The supernatant was discarded, and 500 µL
   of cold acetone was added to each sample. These were then centrifuged
   at 16,000g, 4 °C for 5 min. The supernatant was discarded and replaced
   with 500 µL of cold 70% ethanol (mass spectrometry grade), followed by
   centrifugation at 16,000 g for 5 min at 4 °C. Finally, the supernatant
   was discarded and replaced with 500 µL of cold acetone (mass
   spectrometry grade), followed by centrifugation at 16,000 g, 4 °C for 5
   min. After the supernatant was discarded, the sample was placed in a
   suspension dryer to remove any remaining solvent. The following day,
   each sample was treated with 50 µL of 8 mm UA buffer, gently agitated
   with a test tube rack several times, and placed in a mixer at 25 °C,
   1300 rpm for 2 h. After centrifugation, each sample was alkylated with
   IAA in a mixer at 37 °C, 1300 rpm for 2 h. Trypsin (typically 100 µg,
   corresponding to 2 µL of enzyme) was added to the samples and placed in
   a mixer at 37 °C, 1300 rpm for 16 h. Subsequently, the samples
   underwent desalting, vacuum suspension, and peptide concentration
   determination using a filter solution. Finally, the samples were
   subjected to secretomics analysis using an Orbitrap Exploris 480 mass
   spectrometer (ThermoFisher Scientific) with a data‐dependent
   acquisition mode. The Proteome Discoverer software suite (Version 2.3,
   ThermoFisher Scientific) was utilized to perform peptide identification
   and quantification of the samples, as previously described. For
   proteomics analysis of cancer cells, the main experimental process
   remained consistent with the secretomics analysis. The distinctions in
   sample collection and preliminary treatment are summarized as follows:
   supernatants were discarded, cells were washed twice with PBS, protein
   lysate was added, and proteins were scraped with a cell scraper.
   Centrifugation was conducted at 16,000 g at 4 °C for 10 min. The
   obtained protein supernatant was transferred to a new EP tube, and the
   protein concentration was determined. Subsequently, 200–300 µg of
   protein was used for downstream mass spectrometry sample preparation,
   involving the addition of four times the volume of isopropyl alcohol.
   The mixture was mixed and then placed in −80 °C for precipitation for
   1–2 h. The subsequent procedures were carried out with the supernatant
   samples.

   For secretomics analysis comparing conditioned medium from primary
   cancer cells with that from paired pericytes (PCs), both were isolated
   from tumor tissues of patients with confirmed EGFR mutations. Briefly,
   fresh tumor tissues were washed in saline solution to remove blood.
   Specimens were micro‐dissected into small pieces (0.5–1 cm^3). The
   tissues were enzymatically dissociated at 37 °C for 45 min in a shaking
   incubator using 5 mL of serum‐free DMEM containing 1 mg mL^−1
   collagenase type 3 (Worthington‐#[227]LS004182) and 1 mg mL^−1 DNase I
   (Roche‐#10104159). Digestion was halted by adding an equal volume of
   DMEM with 10% FBS. The cell suspension was then filtered sequentially
   through 100 and 40 µm strainers. Cells retained by the 40 µm strainer
   were resuspended in culturing medium, seeded into 6‐well plates, and
   cultured undisturbed for 5–7 days. Filtered cells were treated with RBC
   Lysis Buffer (BioLegend‐#420301) to remove red blood cells, and cancer
   cells were purified using magnetic‐activated cell sorting (MACS). The
   tissue pellet was resuspended in 100 µL of MACS separation buffer and
   incubated with 10 µL of anti‐EpCAM microbeads (Miltenyi
   Biotec‐#130‐061‐101). EpCAM‐positive cells were magnetically sorted
   using a MACS column (Miltenyi Biotec‐#130‐042‐401), resuspended in DMEM
   with 10% FBS, 10 ng mL^−1 bFGF, 10 ng mL^−1 EGF and 10 ng mL^−1 IGF,
   and cultured in 6‐well plates in a humidified incubator at 37 °C with
   5% CO[2]. Additionally, the negative sorted cells were used for PC
   isolation as previously described.^[ [228]^12b ^] Conditioned medium
   from primary EGFR‐mutated cancer cells or paired PCs was then harvested
   for secretomics analysis.

Statistical Analysis

   Statistical analyses in this study were conducted using several
   methods, including unpaired two‐tailed Student's t‐tests, two‐sided
   Pearson correlation coefficients, log‐rank (Mantel‐Cox) tests, and
   One‐/Two‐way ANOVA followed by Tukey's post hoc testing, depending on
   the data type. The specific statistical tests and exact sample sizes
   (n) are detailed in the figure legends. All in vitro experiments were
   independently replicated at least three times. Statistical calculations
   were performed using GraphPad Prism, with data expressed as
   means ± SEM. Statistical significance was defined as a p‐value of
   < 0.05.

Data Availability

   The mRNA sequencing data and clinicopathological information of NSCLC
   patients were obtained from the website
   ([229]https://kmplot.com/analysis/index.php?p=service&cancer=lung),
   following the guidelines provided by the respective platforms and as
   described previously.^[ [230]^40 ^] The proteomics/secretomics data
   generated in the study had been deposited in the Proteome X consortium
   via the PRIDE partner repository under accession codes (PXD048904,
   PXD048950, PXD055731). No new code was created for this study. The data
   supporting the findings of this study are available in the article and
   supplementary materials.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   C.H., X.H., X.Q., and X.K. contributed equally to this work. C.H.,
   X.H., X.Q., and X.K. conducted the majority of experiments and
   contributed equally to the paper. C.H., X.H., X.Q., C.L., and X.K.
   performed animal experiments, clinical data, and sample analysis. C.H.,
   X.H., C.W., X.J., and C.L. conducted proteomics and secretomics
   experiments and analyzed the data. C.H., X.H., C.W., and L.S. performed
   co‐culture and conditioned medium experiments. M.Y. and M.W. provided
   clinical samples and technical guidance. M.W., L.S., C.L., and P.‐P.W.
   co‐conceived the study, supervised the research, and co‐wrote the
   manuscript. All authors contributed to drafting and approving the final
   manuscript version.

Supporting information

   Supporting Information
   [231]ADVS-11-2405130-s001.docx^ (7.4MB, docx)

Acknowledgements