Abstract

   During targeted treatment, oncogene-addicted tumor cells often evolve
   from an initial drug-sensitive state through a drug-tolerant persister
   bottleneck toward the ultimate emergence of drug-resistant clones. The
   molecular basis underlying this therapy-induced evolutionary trajectory
   has not yet been completely elucidated. Here, we employed a
   multifaceted approach and implicated the convergent role of
   transcription factor Yin Yang 1 (YY1) in the course of diverse targeted
   kinase inhibitors. Specifically, pharmacological perturbation of the
   receptor tyrosine kinase (RTK)/mitogen-activated protein kinase (MAPK)
   pathway resulted in the downregulation of YY1 transcription, which
   subsequently resumed upon therapeutic escape. Failure to decrease YY1
   subverted cytotoxic effects, whereas elimination of residual YY1
   maximized anticancer efficacy and forestalled the emergence of drug
   resistance. Mechanistically, YY1 was uncovered to dictate cell cycle
   and autophagic programs. Immunohistochemical analysis on a wide
   spectrum of clinical specimens revealed that YY1 was ubiquitously
   expressed across lung adenocarcinomas and exhibited anticipated
   fluctuation in response to corresponding RTK/MAPK inhibition. These
   findings advance our understanding of targeted cancer management by
   highlighting YY1 as a determinant node in the context of
   genotype-directed agents.

   Subject terms: Targeted therapies, Transcriptional regulatory elements

Background

   Stepwise progression to drug resistance represents a major challenge in
   anticancer treatment targeting driver oncoproteins such as EGFR, ALK,
   KRAS, BRAF, and many others [[60]1]. Initially, presumed lethal
   therapeutic insults often yield dramatic clinical responses.
   Nevertheless, a reservoir of drug-tolerant persister cells, commonly
   termed minimal residual disease, invariably manage to survive and
   eventually result in tumor relapse [[61]2, [62]3]. Throughout this
   evolutionary process, an increasing variety of genetic [[63]4, [64]5],
   epigenomic [[65]6, [66]7], transcriptional [[67]8, [68]9],
   translational [[69]10, [70]11], post-translational [[71]12, [72]13],
   and metabolic [[73]14, [74]15] mechanisms operate to underpin the
   transition and maintenance of discrete stages, depending on specific
   cancers and therapies. Such a multifactorial and heterogeneous
   regulatory landscape poses significant obstacles to effective
   management and durable control of recalcitrant malignancies [[75]16].
   Further in-depth understanding of molecular profiles at different
   layers and phases may provide new insights into convergent events
   underlying how neoplastic cells respond and evade targeted agents.

   Chromatin accessibility dynamics perhaps constitute one of the most
   relevant but overlooked genomic characteristics corresponding to the
   meticulous changes following treatment with targeted therapies [[76]17,
   [77]18]. The phenotypic alterations associated with the sequential
   drug-sensitive, drug-tolerant, and drug-resistant switch probably
   require widespread remodeling of open chromatin architecture and
   consequent rewiring of gene expression programs. Indeed, mounting
   evidence implicates unique chromatin states in the response, survival,
   and resurgence of cancer cells that are determined by varied epigenetic
   modifiers [[78]6, [79]19–[80]23]. However, additional key players
   involved in this process remain to be elucidated for developing
   rational therapeutic strategies.

   Here, we integrated the assay for transposase-accessible chromatin
   using sequencing (ATAC-seq) and transcriptome-wide RNA-sequencing
   (RNA-seq) to identify Yin Yang 1 (YY1) as a pivotal transcription
   factor in the context of diverse molecular therapeutics targeting
   receptor tyrosine kinase (RTK)/mitogen-activated protein kinase (MAPK).
   YY1 was transcriptionally downstream of RTK/MAPK signaling and governed
   the anticancer effects of pathway blockade by regulating cell cycle and
   autophagic flux. These mechanistic findings illuminate a general hub
   underlying drug-induced evolutionary trajectories in oncogene-addicted
   tumor cells, which holds immense promise for novel intervention
   opportunities.

Results

Characterization of chromatin accessibility dynamics implicates YY1 in EGFR
inhibitor response

   We modeled the drug-sensitive (DS), drug-tolerant (DT), and
   drug-resistant (DR) continuum by using the clinically approved EGFR
   inhibitor erlotinib to treat a well-studied EGFR-mutant lung cancer
   cell line PC9, which was subjected to ATAC-seq and RNA-seq in parallel
   (Fig. [81]1A). Principal component analysis unveiled a comparable
   pattern of these different states in ATAC-seq and RNA-seq results, with
   DT cells being clearly separated from DS and DR cells (Fig. [82]1B).
   The above observation was confirmed by volcano plots of differential
   chromatin accessibility (Supplementary Fig. [83]1A) and gene expression
   (Supplementary Fig. [84]1B). ATAC-seq profiling identified seven
   distinctive clusters of accessible DNA loci (Supplementary Fig.
   [85]2A), which were presumably occupied by diverse transcription
   factors (Supplementary Fig. [86]2B). In addition, we performed motif
   enrichment analysis on the entire set of ATAC-seq peaks to infer
   candidate functional transcription factors and filtered them by
   querying the corresponding RNA-seq data (Fig. [87]1C). Based on this
   integrated approach, a group of transcription factors displaying
   dynamic changes throughout targeted treatment were selected for a
   focused CRISPR-Cas9 screen. YY1, an evolutionarily conserved
   GLI-Krüppel zinc finger transcription factor [[88]24], conferred cell
   suppression upon genetic depletion (Fig. [89]1D). We leveraged the
   CUT&Tag technology (Supplementary Fig. [90]3A) to verify that
   genome-wide YY1 binding underwent a marked decrease in DT cells
   compared to DS cells and was completely restored in DR cells (Fig.
   [91]1E). Of note, the gene expression of various putative cancer stem
   cell markers exhibited inconsistent changes (Supplementary Fig.
   [92]3B), suggesting a limited role for cancer stemness. Therefore, YY1
   might partially account for kinetic alterations in chromatin openness
   and serve as a potential regulator of the EGFR inhibitor response.

Fig. 1. Characterization of chromatin accessibility dynamics implicates YY1
in EGFR inhibitor response.

   [93]Fig. 1
   [94]Open in a new tab

   A A schematic diagram illustrates the modeling of therapy-induced
   evolutionary trajectory in PC9 cells and integrated sequencing analyses
   to identify candidate transcription factors involved in EGFR inhibitor
   response. B Principal component analysis of ATAC-seq and RNA-seq data
   derived from PC9 cells at the initial drug-sensitive state (DS),
   drug-tolerant persister state (DT), deliberate drug-withdrawn state
   (DW), and acquired drug-resistant state (DR). C Heatmaps showing
   dynamic changes of binding motif enrichment (left) and gene expression
   (right) for candidate transcription factors involved in EGFR inhibitor
   response. Genes labeled in red displaying upregulation from DS to DT
   and genes labeled in green displaying upregulation from DT to DR were
   selected for subsequent experiments. D Candidate transcription factors
   involved in EGFR inhibitor response were each knocked out in parental
   PC9 cells or PC9 cells with EGFR^T790M mutant, and cell viability was
   assayed by crystal violet staining. Scale bar represents 10 mm. E
   CUT&Tag profiles showing dynamic changes of YY1 occupancy in PC9 cells
   at the initial drug-sensitive state (DS), drug-tolerant persister state
   (DT), and acquired drug-resistant state (DR).

EGFR regulates YY1 expression through the MAPK signaling pathway

   Correlated with fluctuating YY1 occupancy following EGFR inhibitor
   treatment, YY1 protein abundance notably declined in DT cells relative
   to DS or DR cells, which was ascribable to altered YY1 gene
   transcription (Fig. [95]2A). Corroborating the pharmacological
   evidence, genetic EGFR depletion likewise resulted in YY1
   downregulation at both mRNA and protein levels, whereas YY1 ablation
   did not affect EGFR expression (Fig. [96]2B). Moreover, immunoblotting
   assays revealed that YY1 intensity fully emulated EGFR phosphorylation
   when PC9 cells were exposed to the first-, second-, or third-generation
   EGFR inhibitors in a time course fashion (Fig. [97]2C). Similar
   associations were recapitulated in various EGFR-mutant lung cancer cell
   lines including HCC827, NCI-H1975, and NCI-H820 (Supplementary Fig.
   [98]4A). Immunofluorescence staining indicated predominant nuclear
   localization of YY1, which showed diminished signals consequent to EGFR
   blockade (Fig. [99]2D). MAPK and PI3K signaling cascades represented
   two major pathways downstream of EGFR kinase to possibly regulate YY1
   expression. We sought to dissect their specific contributions by
   administering trametinib (an MEK inhibitor), ulixertinib (an ERK
   inhibitor), pictilisib (a PI3K inhibitor), or ipatasertib (an AKT
   inhibitor) to block either axis (Fig. [100]2E). In PC9 cells, applying
   trametinib or ulixertinib was sufficient to induce YY1 downregulation
   while pictilisib or ipatasertib had no equivalent effects (Fig.
   [101]2F), supporting an exclusive role of MAPK but not PI3K pathway.
   Such a notion was substantiated by validation experiments in HCC827,
   NCI-H1975, and NCI-H820 cells (Supplementary Fig. [102]4B). Analogous
   to erlotinib, trametinib also attenuated YY1 mRNA expression
   (Supplementary Fig. [103]4C). Thus, our data unambiguously proved that
   EGFR regulated YY1 expression through the MAPK signaling pathway.
   Within the MAPK pathway, we further tested a number of candidate
   transcription factors that were predicted to bind to the YY1 promoter,
   including ETS1, CREB1, SRF, ELK1, ETV1, SP1, EGR1, RREB1, MAZ, and
   CEBPA. However, none of them showed notable regulatory effects on YY1
   protein levels based on siRNA knockdown experiments (Supplementary Fig.
   [104]4D). Given the multitude and complexity of transcription factors
   downstream of the MAPK pathway, larger-scale screening might be needed,
   or alternatively, multiple transcription factors could compensate for
   each other to control YY1 expression.

Fig. 2. EGFR regulates YY1 expression through the MAPK signaling pathway.

   [105]Fig. 2
   [106]Open in a new tab

   A Relative YY1 protein and YY1 mRNA levels in PC9 cells at the initial
   drug-sensitive state (DS), drug-tolerant persister state (DT), and
   acquired drug-resistant state (DR). B EGFR or YY1 gene was knocked out
   in PC9 cells using the CRISPR-Cas9 system. The indicated proteins were
   analyzed by immunoblotting, and the relative expression of YY1 mRNA was
   measured by quantitative PCR analysis. C Parental PC9 cells or PC9
   cells with EGFR^T790M mutant were treated with erlotinib, afatinib or
   osimertinib in a time course manner, and the indicated proteins were
   analyzed by immunoblotting. D Immunofluorescence staining of YY1 (red)
   in parental PC9 cells or PC9 cells with EGFR^T790M mutant in the
   presence of erlotinib or osimertinib treatment, respectively. Cell
   nuclei were counterstained with DAPI (blue). Scale bar represents
   50 μm. E A schematic diagram illustrating the EGFR signaling pathway
   and corresponding kinase inhibitors that may modulate YY1 expression. F
   Parental PC9 cells (DS) or PC9 cells with EGFRT790M mutant (DR) were
   treated with MAPK pathway inhibitors (trametinib and ulixertinib) or
   PI3K pathway inhibitors (pictilisib and ipatasertib) in a time course
   manner, and the indicated proteins were analyzed by immunoblotting.

RTK/MAPK inhibition downregulates YY1 in diverse oncogene-addicted cancer
models

   We reasoned that YY1 function might not be restricted to EGFR-mutant
   lung cancer. Indeed, YY1 gene expression was unrelated to EGFR status
   in lung adenocarcinoma from the Cancer Genome Atlas (Fig. [107]3A). In
   addition, higher levels of YY1 mRNA tended to predict inferior patient
   survival irrespective of EGFR alterations (Fig. [108]3B). At the
   protein level, immunoblotting assays illustrated that YY1 was readily
   detectable in a wide spectrum of oncogene-driven lung cancer cell lines
   (Fig. [109]3C), as well as in most primary tumor samples regardless of
   EGFR mutation (Fig. [110]3D). Based on these observations, a panel of
   lung cancer cell lines bearing heterogeneous RTK/MAPK aberrations was
   assembled and treated with corresponding kinase inhibitors. YY1
   abundance was appreciably decreased along with ERK repression in a
   time-dependent manner (Fig. [111]3E). We then extended our
   investigation to other cancer types with RTK/MAPK pathway
   dysregulation. Two examples, erdafitinib (FGFR inhibitor) and sotorasib
   (KRAS G12C inhibitor) impaired YY1 expression in genotype-matched
   bladder cancer cell line RT112 and pancreatic cancer cell line MIA
   PaCa-2, respectively (Fig. [112]3F). We concluded that RTK/MAPK
   inhibition downregulated YY1 in diverse oncogene-addicted cancer
   models.

Fig. 3. RTK/MAPK inhibition downregulates YY1 in diverse oncogene-addicted
cancer models.

   [113]Fig. 3
   [114]Open in a new tab

   A Violin plots indicating the relative expression of YY1 mRNA in TCGA
   lung adenocarcinomas (LUAD) with or without EGFR mutations. B
   Kaplan–Meier plots illustrating associations between YY1 expression and
   overall survival in 80 TCGA LUAD patients with EGFR mutations (top) and
   427 TCGA LUAD patients without EGFR mutations (bottom). P-values are
   based on log-rank tests. The light red or blue shading indicates 95%
   confidence interval. C YY1 protein was analyzed by immunoblotting in a
   panel of lung cancer cell lines with diverse oncogenic alterations. D
   YY1 protein was analyzed by immunoblotting in normal lung tissues and
   paired tumor lesions from human lung adenocarcinomas with or without
   EGFR mutations. E Lung cancer cell lines with diverse oncogenic
   alterations were treated with genotype-matched kinase inhibitors in a
   time course manner, and the indicated proteins were analyzed by
   immunoblotting. F A bladder cancer cell line with FGFR3 rearrangement
   and a pancreatic cancer cell line with KRAS mutation were treated with
   genotype-matched kinase inhibitors in a time course manner, and the
   indicated proteins were analyzed by immunoblotting.

YY1 depletion underlies therapeutic effects of molecular targeted agents

   We wondered whether YY1 exerted functional activities, especially in
   the context of molecular targeted regimens. Initially, lentiviral
   CRISPR-Cas9 technology was employed to knockout YY1 in multiple
   EGFR-mutant, KRAS-altered, or ALK-rearranged lung cancer cell lines
   using two independent sgRNAs (Fig. [115]4A), and cell viability was
   significantly inhibited without exception (Fig. [116]4B). Subsequently,
   we genetically manipulated YY1 expression at discrete phases of EGFR
   inhibitor treatment and assessed its impact on therapeutic outcomes
   (Fig. [117]4C). First, YY1 ablation in DS cells improved drug
   sensitivity by preventing DT emergence in PC9 (Fig. [118]4D), as well
   as HCC827, NCI-H1975, and NCI-H820 (Supplementary Fig. [119]5A).
   Second, YY1 overexpression in DS cells antagonized the drug response by
   accelerating DT formation in PC9 (Fig. [120]4E), as well as HCC827,
   NCI-H1975, and NCI-H820 (Supplementary Fig. [121]5B). Third, YY1
   knockout in DT cells, leading to drastically compromised proliferation
   estimated by EdU incorporation assays (Fig. [122]4F), eliminated
   residual colonies and overcame DR expansion (Fig. [123]4G). Finally, we
   conducted in vivo experiments to confirm the in vitro findings. Upon
   YY1 loss, PC9 xenograft growth was slowed and more susceptible to
   erlotinib inhibition (Fig. [124]4H), yielding reduced tumor masses
   (Fig. [125]4I) with less weight (Fig. [126]4J). In contrast, exogenous
   YY1 expression attenuated erlotinib efficacy (Fig. [127]4K), as
   exemplified by larger tumor masses (Fig. [128]4L) with gained weight
   (Fig. [129]4M). Collectively, these studies suggested that YY1
   depletion underlay the therapeutic effects of molecular targeted
   agents.

Fig. 4. YY1 depletion underlies the therapeutic effects of molecular targeted
agents.

   [130]Fig. 4
   [131]Open in a new tab

   A YY1 was knocked out in EGFR-mutant, KRAS-altered, or ALK-rearranged
   lung cancer cell lines with two independent sgRNAs using the
   CRISPR-Cas9 system, and YY1 protein was analyzed by immunoblotting. B
   YY1 was knocked out in EGFR-mutant, KRAS-altered, or ALK-rearranged
   lung cancer cell lines with two independent sgRNAs using the
   CRISPR-Cas9 system, and cell viability was assessed by crystal violet
   staining. Scale bar represents 10 mm. C A schematic diagram illustrates
   the genetic perturbation of YY1 or EGFR in PC9 cells at different
   states. Top and bottom: YY1 was knocked out or overexpressed in PC9
   cells at the initial drug-sensitive state (DS) and then cells were
   treated with erlotinib to induce drug-tolerant persister state (DT);
   middle, PC9 cells were treated with erlotinib to form drug-tolerant
   persister state (DT), and YY1 or EGFR was manipulated as indicated to
   track acquired drug-resistant state (DR). D YY1 was knocked out in PC9
   cells, and YY1 protein was analyzed by immunoblotting. EGFR inhibitor
   response upon YY1 depletion was assessed by crystal violet staining.
   Scale bar represents 10 mm. E YY1 was overexpressed in PC9 cells, and
   YY1 protein was analyzed by immunoblotting. EGFR inhibitor response
   upon YY1 overexpression was assessed by crystal violet staining. Scale
   bar represents 10 mm. F YY1 was knocked out in PC9-derived
   drug-tolerant (DT) cells with (left) or without (right) EGFR^T790M
   mutant, followed by immunofluorescence staining of YY1 (red) and EdU
   (green). Cell nuclei were counterstained with DAPI (blue). Scale bar
   represents 50 µm. G YY1 was knocked out in PC9-derived drug-tolerant
   (DT) cells with (left) or without (right) EGFR^T790M mutant and cell
   viability was assessed by crystal violet staining. Scale bar represents
   10 mm. H Tumor growth curves of YY1-depleted PC9 xenografts that were
   treated with vehicle or erlotinib. Each line represents the mean tumor
   volume of the respective group. Error bars indicate standard deviation
   (8 mice/group). *P < 0.05, unpaired Student’s t-test. I Representative
   images of PC9 xenografts with or without YY1 knockout. The xenografts
   were treated with vehicle (30% PEG 300, 5% Tween 80) or erlotinib
   (5 mg/kg/day). Scale bar represents 10 mm. J Tumor weights of
   YY1-depleted PC9 xenografts that were treated with vehicle or
   erlotinib. Error bars indicate standard deviation (8 mice/group).
   *P < 0.05, unpaired Student’s t-test. K Tumor growth curves of
   YY1-overexpressed PC9 xenografts that were treated with vehicle or
   erlotinib. Each line represents the mean tumor volume of the respective
   group. Error bars indicate standard deviation (8 mice/group).
   *P < 0.05, unpaired Student’s t-test. NS denotes not significant. L
   Representative images of PC9 xenografts with or without YY1
   overexpression. The xenografts were treated with vehicle (30% PEG 300,
   5% Tween 80) or erlotinib (5 mg/kg/day). Scale bar represents 10 mm. M
   Tumor weights of YY1- overexpressed PC9 xenografts that were treated
   with vehicle or erlotinib. Error bars indicate standard deviation
   (8 mice/group). *P < 0.05, unpaired Student’s t-test. NS denotes not
   significant.

YY1 dictates cell cycle and autophagic programs

   To determine the mechanistic underpinnings of YY1 function, RNA-seq was
   performed to compare YY1-null and parental PC9 cells (Supplementary
   Fig. [132]6A). As a transcription factor, YY1 knockout resulted in a
   considerable number of differentially expressed genes (DEGs) (Fig.
   [133]5A; Supplementary Fig. [134]6B). There was a significant positive
   correlation between fold changes induced by two sgRNAs at the
   transcriptome-wide scale (Supplementary Fig. [135]6C), and Venn diagram
   illustrated that 2835 out of 5811 DEGs (48.8%) were shared
   (Supplementary Fig. [136]6D). Gene ontology annotation pinpointed that
   cell cycle-related terms were downregulated while stress- and
   autophagy-related pathways were upregulated (Fig. [137]5B). Gene set
   enrichment analysis (Fig. [138]5C) and Metascape algorithm
   (Supplementary Fig. [139]7A) validated the enriched biological
   processes associated with YY1 ablation. As the experimental proof, cell
   cycle examination by flow cytometry uncovered that YY1-depleted PC9
   cells were arrested at the G2/M phase (Fig. [140]5D). A closer
   inspection on the autophagic machinery found that in the absence of YY1
   [[141]25, [142]26], many regulators were widely altered (Supplementary
   Fig. [143]7B), among which early autophagy-related genes were generally
   upregulated whereas late autophagy-related genes were frequently
   downregulated (Supplementary Fig. [144]7C). LC3-II, a commonly used
   autophagy indicator, was markedly accumulated in PC9 (Fig. [145]5E), as
   well as NCI-H1975, NCI-H23, and NCI-H3122 (Fig. [146]5F). Of note,
   similar phenomenon was observed in chloroquine-treated PC9 cells, in
   line with autophagic flux suppression. Consistently, the tandem
   RFP-GFP-LC3 probe revealed a rise in autophagosomes (RFP^+GFP^+ puncta)
   rather than acidified autolysosomes (RFP^+GFP^- puncta) (Fig. [147]5G).
   Morphometric transmission electron microscopy identified abundant
   autophagosomes but few autolysosomes upon YY1 depletion, suggestive of
   deficient autophagosome-lysosome fusion (Fig. [148]5H). These data
   supported that YY1 might affect therapeutic response to molecular
   targeted agents by dictating cell cycle and autophagic programs
   [[149]27–[150]31].

Fig. 5. YY1 dictates cell cycle and autophagic programs.

   [151]Fig. 5
   [152]Open in a new tab

   A Volcano plot showing differentially expressed genes after YY1
   knockout by two independent sgRNAs in PC9 cells. B Gene ontology
   categories overrepresented in differentially expressed genes upon YY1
   knockout. Red arrows indicate upregulated pathways and blue arrows
   indicate downregulated pathways. Q value represents the adjusted P
   value. C Enrichment plots from gene set enrichment analysis (GSEA) of
   YY1-related (red), cell cycle-related (blue), and autophagy-related
   (green) gene signatures. All GSEA results were statistically
   significant (FDR < 0.05). D Cell cycle analysis of YY1-depleted PC9
   cells by flow cytometry and bar graphs showing the distribution of the
   cell cycle phases. E YY1 gene was knocked out in PC9 cells, and the LC3
   protein was analyzed by immunoblotting in the presence or absence of
   chloroquine, a late-stage autophagy inhibitor. F YY1 gene was knocked
   out in NCI-H1975, NCI-H23, or NCI-H3122 cells, and the LC3 protein was
   analyzed by immunoblotting. G YY1-depleted PC9 cells were transfected
   with the RFP-GFP-LC3 plasmid and analyzed by confocal microscopy.
   Yellow puncta indicate autophagosomes while red puncta indicate
   autolysosomes. The top panel shows the quantification of LC3 puncta per
   cell. *P < 0.05, unpaired Student’s t-test. Scale bar represents 10 µm.
   H Representative images of transmission electron microscopy
   demonstrating abnormal accumulation of autophagosomes and autolysosomes
   in YY1-depleted PC9 cells. Scale bar of the upper images represents
   1 µm and the scale bar of the lower images represents 0.2 µm.

YY1 exhibits anticipated fluctuation following targeted treatment in lung
cancer patients

   To investigate the clinical relevance of YY1 expression, three cohorts
   of lung adenocarcinoma were collected. First, immunohistochemical
   interrogation of 89 EGFR-mutant, 51 KRAS-altered, and 46 ALK-rearranged
   lung cancer specimens unveiled that YY1 was ubiquitously expressed in
   malignant cells at an elevated level relative to the normal epithelium
   (Fig. [153]6A; Supplementary Table [154]1). Second, YY1 was evaluated
   in four lung cancer patients who harbored somatic EGFR mutations and
   received off-label neoadjuvant targeted therapies (Fig. [155]6B;
   Supplementary Table [156]2). Although YY1 was highly expressed in tumor
   lesions compared to adjacent normal tissues at baseline, a consistent
   reduction was observed in post-treatment versus pre-treatment samples
   (Fig. [157]6C). Third, YY1 was analyzed in six cases of advanced
   EGFR-mutant lung adenocarcinoma with disease relapse following
   first-line EGFR inhibitor treatment (Fig. [158]6D; Supplementary Table
   [159]3), which showed that YY1 levels rebounded to baseline in
   recurrent tumors (Fig. [160]6E). Taken together, YY1 was ubiquitously
   expressed and exhibited anticipated fluctuation following targeted
   treatment in lung cancer patients.

Fig. 6. YY1 exhibits anticipated fluctuation following targeted treatment in
lung cancer patients.

   [161]Fig. 6
   [162]Open in a new tab

   A Immunohistochemistry staining of YY1 in a cohort of 186 patients
   affected by EGFR-mutant, KRAS-altered, or ALK-rearranged lung cancer.
   Blue dots indicate normal lung tissues, and orange dots indicate paired
   tumor lesions. P-values are based on paired Student’s t-tests. Scale
   bar represents 50 µm. B Computed tomography (CT) images and
   representative YY1 immunohistochemistry staining in EGFR-mutant lung
   adenocarcinomas before and after neoadjuvant EGFR inhibitors. The
   H-score system was used for immunohistochemical quantification. Scale
   bar represents 50 µm. C Dot plots indicating intensity changes of YY1
   staining in normal lung tissues and paired tumor lesions before and
   after neoadjuvant targeted therapy. D Representative YY1
   immunohistochemistry staining in EGFR-mutant lung adenocarcinomas at
   baseline and upon disease recurrence following first-line EGFR
   inhibitor treatment. The H-score system was used for
   immunohistochemical quantification. Scale bar represents 50 µm. E Dot
   plots indicating intensity changes of YY1 staining in normal lung
   tissues and paired tumor lesions at baseline and upon disease
   recurrence following first-line EGFR inhibitor treatment.

Discussion

   This study presented a global map of chromatin accessibility and
   transcriptomic landscape during EGFR perturbation. Integrated
   bioinformatics and functional analyses highlighted a model implicating
   the transcription factor YY1 as a master regulator downstream of
   RTK/MAPK signaling. We showed how YY1 might govern antineoplastic
   output by dictating cell cycle and autophagic programs. These
   discoveries revealed an exciting and previously unappreciated component
   in charge of therapeutic response to molecular targeted agents and
   pointed to future avenues for improving the magnitude and duration of
   clinical benefit.

   Despite the proven significance and current limitations of targeted
   treatment to reduce cancer mortality, our understanding of
   therapy-induced evolutionary trajectory remains surprisingly incomplete
   [[163]32]. Recent literature has individually focused on persister or
   refractory cells, both constituting major causes of disease relapse
   [[164]2, [165]3, [166]16]. Using EGFR-mutant lung cancer as a
   representative system, we extended prior investigations in two
   dimensions to address this unsolved fundamental challenge. First, the
   whole progressive course upon drug intervention was modeled across the
   continuum of drug-sensitive, drug-tolerant, and drug-resistant phases
   to approximate a typical scenario in clinic. Second, ATAC-seq and
   RNA-seq were combined to thoroughly profile the molecular dynamics at
   different layers. Such a comprehensive interrogation allowed to
   identify that YY1 underwent kinetic alterations of genomic occupancy
   and gene expression. Of special importance, YY1 was found to be not
   only modulated by EGFR but also broadly controlled by RTK/MAPK
   signaling at the transcriptional level, although the downstream
   transcription factors require continued research. Thus, our research
   pinpointed a universal program operating in the course of diverse
   kinase inhibitors targeting the RTK/MAPK cascade to treat
   oncogene-driven malignancies. It is noteworthy that YY1 could not be
   fully eliminated even with prolonged drug exposure, implying that
   additional regulatory modes may exist and warrant further exploration.

   YY1 belongs to the GLI-Krüppel family of zinc finger transcription
   factors [[167]24]. It is highly expressed in many tumor types and
   regulates numerous genes involved in hallmarks of cancer including
   uncontrolled cell proliferation, invasive behavior, metabolic
   reprogramming, and drug resistance [[168]33–[169]39]. Along this line,
   our data unraveled that cell cycle and autophagic flux, among a
   plethora of other pathways, were evidently impacted by YY1 ablation,
   which offered mechanistic interpretations of YY1 function in anticancer
   treatment. Moreover, YY1 can act as a transcriptional activator or
   repressor depending on the context. Indeed, we discovered that YY1
   exerted dichotomous effects on early or late autophagy-related genes,
   aggregately leading to defective autophagic progression. Beyond its
   role as a traditional transcription factor, YY1 has been reported to
   promiscuously interact with cofactors and chromatin modifiers
   [[170]40], mediate enhancer-promoter looping [[171]41], and possibly
   form phase-separated condensates [[172]42]. Therefore, its exact
   mechanism of action underlying targeted therapeutics needs future
   in-depth dissection.

   One critical question concerns whether preclinical observations
   recapitulate the clinical reality in cancer patients. Notably, we
   collected three cohorts of lung adenocarcinoma to resolve this issue.
   At baseline, YY1 was ubiquitously detected across a total of 186
   primary tumors spanning EGFR-mutant, KRAS-altered, and ALK-rearranged
   specimens. In four EGFR-mutant subjects receiving EGFR inhibitors and
   longitudinal sampling, there was a consistent decrease of YY1 protein
   in paired post-treatment versus pre-treatment sections. Finally, YY1
   expression was invariably resumed upon disease recurrence following
   first-line EGFR inhibitor treatment in six EGFR-mutant patients.
   Collectively, we demonstrated that YY1 levels determined the cytotoxic
   response to molecular targeted agents, providing a promising biomarker
   for evaluating drug effectiveness. In addition, our work establishes a
   sound rationale to develop combination regimens that eradicate YY1
   remnants at drug-tolerant states or prohibit YY1 recovery during
   drug-resistant transit. Considering that oncogenic activation of the
   EGFR pathway results in an immune-inert phenotype and recent literature
   has implicated a potential role of YY1 in mediating tumor immune escape
   [[173]36, [174]43–[175]46], these conclusions may also be tentatively
   extrapolated to cancer immunotherapy.

Materials and methods

Cell culture and reagents

   Tumor cell lines and HEK293T cells were originally obtained from the
   American Type Culture Collection (ATCC) or Japanese Collection of
   Research Bioresources Cell Bank (JCRB), where mycoplasma contamination
   was routinely tested and cell identity was monitored by short tandem
   repeat (STR) profiling. Drug-tolerant persister cells were generated by
   continuous treatment with 1 μM erlotinib for 10 days, according to
   protocols described previously [[176]6]. Drug-resistant cells were
   generated by overexpressing the EGFR^T790M mutant in PC9 cells and then
   maintained in 1 μM erlotinib. Cells were cultured in RPMI1640 (Life
   Technologies) supplemented with 10% fetal bovine serum (Gibco),
   l-glutamine (2 mM), penicillin (100 units/mL), and streptomycin
   (100 μg/mL). Small-molecule inhibitors were purchased from Selleck
   Chemicals or MedChemExpress and reconstituted in DMSO (Sigma–Aldrich)
   at a stock concentration of 10 mM. The following inhibitors were used
   at these final concentrations unless otherwise indicated: erlotinib
   (1 μM), afatinib (4 μM), osimertinib (4 μM), trametinib (5 μM),
   ulixertinib (1 μM), pictilisib (1 μM), ipatasertib (1 μM), alectinib
   (5 μM), capmatinib (4 μM), PLX8349 (4 μM), erdafitinib (2 μM),
   sotorasib (1 μM), chloroquine (50 μM). For visualization, cells were
   fixed with formalin and stained with crystal violet.

Plasmids, siRNA, and sgRNAs

   Plasmids for gene overexpression were constructed using the Gibson
   Assembly Cloning Kit (New England Biolabs) and Gateway Cloning System
   (Invitrogen). EGFR mutations were generated using the Q5 Site-Directed
   Mutagenesis Kit (New England Biolabs) and verified by Sanger
   sequencing. All siRNAs were used at a final concentration of 25 nM and
   transfected into cells with Lipofectamine 2000 reagent according to the
   manufacturer’s instructions (Thermo Fisher Scientific). The CRISPR-Cas9
   system was employed to knockout indicated genes. The primers used for
   cloning and the sgRNA sequences are provided in Supplementary Table
   [177]4.

Virus production and cell infection

   For virus production, HEK293T cells in a 10-cm dish were co-transfected
   with 5 µg of lentiviral constructs, 5 μg of plasmid Δ8.9, and 3 μg of
   plasmid VSVG using Lipofectamine 2000. Cells were incubated at 37 °C
   and the medium was replaced with fresh complete medium after 12 h.
   Virus-containing medium was collected 48-72 h after transfection and
   supplemented with 8 μg/mL polybrene (Fluka) to infect target cells in
   6-well dishes. Infected cells were selected with 2–5 μg/mL puromycin or
   blasticidin for one week.

Western blotting analysis

   Cells were lysed in RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1%
   NP-40, 0.1% SDS, 2 μM EDTA) containing protease inhibitors (Roche) and
   phosphatase inhibitors (Roche). Protein concentrations were quantified
   using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Cell
   lysates (~20 μg protein) were subjected to SDS-PAGE (Invitrogen) and
   Western blotting. The following primary antibodies were used: YY1
   (ab109237, Abcam), phospho-EGFR (Y1068) (#48576, Cell Signaling
   Technology), EGFR (#4267, Cell Signaling Technology), phospho-ERK
   (T202/Y204) (#9106, Cell Signaling Technology), ERK (#4695, Cell
   Signaling Technology), phospho-AKT (T308) (#2965, Cell Signaling
   Technology), AKT (#2966, Cell Signaling Technology), LC3-II (#12741,
   Cell Signaling Technology), β-actin (#5125, Cell Signaling Technology).

Immunofluorescence staining

   Cells were fixed with 4% paraformaldehyde for 15 min, and permeabilized
   with 0.1% Triton X-100 in PBS for 10 min. After three PBS washes, cells
   were blocked with 2% BSA in PBS for 30 min at room temperature (RT),
   and incubated with primary antibodies against YY1 (#66281-1-Ig,
   Proteintech) diluted in 2% BSA at 4 °C overnight. Cells were incubated
   with Alexa Fluor 488-labeled anti-mouse IgG (A11029, Invitrogen) and
   Alexa Fluor 594-labeled anti-rabbit IgG (A11037, Invitrogen) for 1 h at
   RT in the dark, followed by 4′,6-diamidino-2-phenylindole (DAPI)
   (Invitrogen) counterstaining for 5 min. The staining images were
   acquired using a confocal laser scanning microscope (Leica). For
   RFP-GFP-LC3 analysis, PC9 cells were transfected with 2 μg of tandem
   monomeric RFP-GFP-LC3 plasmid for 48 h and subjected to the indicated
   treatment. For evaluating tandem fluorescent LC3 puncta, the cells were
   rinsed once with PBS followed by confocal microscopy (Leica) analysis.
   The number of LC3 puncta per cell was counted using Image J (v 2.1.0).

EdU incorporation assay

   The EdU (5-ethynyl-2′-deoxyuridine) incorporation assay was performed
   with the Cell-Light^TM EdU DNA Cell Proliferation Kit (RiboBio)
   according to the manufacturer’s protocol. Briefly, cells were treated
   as indicated, incubated with EdU (50 μM) for 2 h, and fixed in 4%
   paraformaldehyde at RT for 30 min. The cells were stained with
   ApolloGreen fluorescent dye, followed by incubation with Hoechst
   reaction solution. Stained samples were viewed under a fluorescence
   microscope (Leica).

Patient samples

   Human samples were obtained in accordance with the ethical guidelines
   of the U.S. Common Rule, and the study was approved by the Ethics
   Committee of Ren Ji Hospital. Written informed consent was acquired
   from all patients. Paired lung adenocarcinoma samples were collected
   before and after neoadjuvant targeted therapy, or at baseline and upon
   disease recurrence following first-line EGFR inhibitor treatment. In
   addition, we assembled two cohorts of treatment-naïve lung
   adenocarcinomas. One cohort contained 18 patients, whose tissue
   specimens were subjected to Western blotting analysis. The other cohort
   contained 186 patients with driver gene mutations for
   immunohistochemical staining of the formalin-fixed and
   paraffin-embedded (FFPE) sections.

Immunohistochemistry staining

   The tissue slides were baked, dewaxed with xylene, passed through
   graded alcohols, and antigen retrieved with 10 mM citric sodium (pH
   6.0) in a steam pressure cooker for 20 min. The slides were then
   treated with 3% hydrogen peroxide solution in methanol for 10 min to
   quench endogenous peroxidase activity, blocked with goat serum, and
   incubated with primary antibodies against YY1 (ab81552, Abcam),
   followed by incubation with horseradish peroxidase-conjugated secondary
   antibody for 1 h at RT. Antigen visualization was performed using
   3,3′-diaminobenzidine (DAB) chromogen (Vector Laboratories). Slides
   were counterstained with hematoxylin, dehydrated, and coverslipped with
   mounting solution (Invitrogen). Whole slides were scanned with a Leica
   Aperio CS2 slide scanner system (Leica Biosystems). The staining was
   assessed independently by two pathologists without knowledge of patient
   characteristics based on the H-score method. The H-score was calculated
   by adding the percentage of positive cells multiplied by an ordinal
   value corresponding to the intensity level (none = 0, weak = 1,
   moderate = 2, strong = 3). H-scores ranged from 0 to 300, and tissue
   samples were further defined as negative (0–49), weak (50–99), medium
   (100–199), or positive (200–300).

Cell cycle analysis

   The cell cycle was measured using iodide (PI) staining and flow
   cytometry analysis. Briefly, genetically edited PC9 cells were
   collected and washed once with PBS. The cells were then fixed with cold
   70% ethanol for 30 min and resuspended in Propidium Iodide (PI)/RNase
   Staining Solution (Cell Signaling Technology). FACS AriaII cytometer
   (BD Biosciences) was used for flow cytometry analysis, and the data
   were processed with FlowJo software.

Transmission electron microscopy

   For transmission electron microscopy (TEM), PC9 cells were treated as
   indicated, washed twice with PBS, and fixed for 6 h at RT in 2.5%
   ultrapure glutaraldehyde in PBS. Post-fixation was performed with 1%
   osmium tetroxide for 90 min at 4 °C, washed four times in PBS, and
   dehydrated with gradient ethanol solutions (Sinopharm Chemical Reagent)
   from 50% to 100% in a 10% graded series. Infiltration was carried out
   using ethanol with propylene oxide (1:1 ratio) for 30 min. The
   specimens were embedded in Epon 812 resins (Ted Pella), followed by
   polymerization at 60 °C for 48 h. Ultrathin sections were cut to a
   thickness of 70 nm by Leica EM UC7 ultramicrotome and stained with
   saturated uranyl acetate in 50% ethanol, followed by Reynolds lead
   citrate. Electron microscopy images were obtained using a TEM system
   (FEI Tecnai G2 Spirit BioTwin).

RNA extraction and quantitative PCR assay

   Total RNA from cells was extracted with RNA Isolation Kit (Vazyme
   Biotech) and converted to cDNA using the High-Capacity cDNA Reverse
   Transcription Kit (Invitrogen). Quantitative real-time PCR was
   performed on a ViiA™ 7 Real-Time PCR System (Applied Biosystems) using
   ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech). All samples were
   normalized to β-actin expression as the endogenous control. At least
   three biological replicates were included for each condition. The
   primer sequences used for PCR are listed in Supplementary Table [178]4.

RNA-sequencing and analysis

   PC9 cells were genetically edited or treated as indicated. Total RNA
   was purified using the RNeasy Plus kit (Qiagen) following the
   manufacturer’s instructions. RNA quality was assessed by NanoDrop 8000
   (Thermo Fisher Scientific) and agarose gel electrophoresis. Sequencing
   libraries were prepared using the NEBNext Ultra RNA Library Prep Kit
   for Illumina (NEB). Clustering of the index-coded libraries on a cBot
   Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS
   (Illumina), and the library preparations were sequenced on an Illumina
   Novaseq platform to generate 30 million 150 bp paired-end reads
   (Novogene). All downstream analyses were based on clean data obtained
   by removing low-quality reads and reads that contained adapters or
   ploy-N sequences. The index of the reference genome was built using
   Bowtie (v2.2.3) [[179]47], and clean reads were aligned to the
   reference genome using HISAT2 (v2.0.5) [[180]48]. FeatureCounts
   (v1.5.0-p3) was used to count the reads numbers mapped to each gene
   [[181]49]. Differential expression analysis was performed using DESeq2
   (v1.20.0) R package [[182]50]. Transcripts with adjusted P-values of
   <0.05 were assigned as differentially expressed genes.

ATAC sequencing

   The assay for transposase-accessible chromatin using sequencing
   (ATAC-seq) was performed as previously described [[183]51]. First, a
   total of 5 × 10^4 cells were harvested in ice-cold washing buffer
   (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl[2], 0.1% IGEPAL CA-630).
   Cell pellets were resuspended in Tn5 transposition reaction mix
   containing 25 µL 2× TD Buffer, 2.5 µL Tn5 Transposes (Illumina), and
   22.5 µL nuclease-free water, and incubated at 37 °C for 30 min. DNA was
   purified with MinElute PCR Purification Kit (Qiagen). Eluted DNA was
   barcoded and amplified using NEBNext Q5 Hot Start HiFi PCR Master Mix
   (#M0543L, New England Labs) in a total volume of 50 µL with the
   following PCR program: 72 °C for 3 min; 98 °C for 30 s; 12 cycles of
   98 °C for 30 s, 63 °C for 30 s and 72 °C for 3 min; 72 °C for 5 min.
   The libraries were purified with QIAquick PCR Purification Kit (Qiagen)
   and sequenced on the Illumina HiSeq with 50-bp paired-ends. Reads were
   aligned to the human reference genome (GRCh37/hg19) using BWA with
   standard parameters [[184]52], and filtered for high-quality
   (MAPQ ≥ 13), non-mitochondrial, and properly paired reads (longer than
   18 nt). MACS2 (v 2.2.6) was used to call peaks using the parameter “-q
   0.05 -g hs --nomodel --shift -75 --extsize 150” [[185]53]. Peaks with
   an FDR of lower than 5% were saved to detect chromosomal regions for
   further analysis. Bigwig files were generated from the filtered BAM
   files using deepTools2 bamCoverage with the option “--normalizeUsing
   RPKM” [[186]54]. Peak annotation was performed with HOMER
   annotatePeaks.pl [[187]55]. Heatmaps and average profile plots were
   generated using deepTools plotHeatmap. Coverage tracks were visualized
   using the trackViewer R package [[188]56].

CUT&Tag analysis

   Cleavage under targets & tagmentation (CUT&Tag) was performed using the
   Hyperactive In-Situ ChIP Library Prep Kit for Illumina (TD901, Vazyme
   Biotech) following the manufacturer’s instructions. Briefly, cells were
   collected into a low absorbing tube, washed twice in 500 μL Wash
   Buffer, and incubated with concanavalin A-coated magnetic beads for
   10 min at RT. The bead-bound cells were resuspended in 50 μL Antibody
   Buffer, followed by incubation with 1 μg of primary antibody at 4 °C
   overnight with slow rotation. The supernatant was removed, and cells
   were incubated with a secondary antibody in a Dig-wash Buffer for 1 h
   at room temperature. Cells were washed three times with Dig-wash Buffer
   and incubated with 0.04 µM Hyperactive pG-Tn5 Transposase adapter mix
   prepared in Dig-300 buffer. Cells were washed three times with Dig-300
   buffer to remove unbound pG-Tn5 and resuspended in 300 µL Tagmentation
   Buffer to incubate for 1 h at 37 °C. The reaction was terminated, and
   DNA was extracted using the phenol-chloroform-isoamyl alcohol method.
   PCR was performed to amplify the libraries under the following cycling
   conditions: 72 °C for 5 min, 98 °C for 30 s, 20 cycles of 98 °C for
   10 s, and 60 °C for 30 s, followed by a final extension at 72 °C for
   1 min and holding at 4 °C. PCR products were then purified with DNA
   Clean Beads. After quantification and quality control, DNA libraries
   were sequenced on an Illumina NovaSeq platform. The following
   antibodies were used: anti-rabbit IgG (#2729, Cell Signaling
   Technology), YY1 (ab109237, Abcam). Peaks were called using SEACR
   (v1.1) and MACS2 (v2.2.6) [[189]53, [190]57]. Subsequent analyses were
   the same as those for the ATAC-seq pipeline.

In vivo studies

   To evaluate the function of YY1 in vivo, 1 × 10^6 PC9 cells with YY1
   genetic editing were mixed with Matrigel (BD Biosciences) and
   subcutaneously implanted in the dorsal flank of BALB/c nude mice (5
   weeks of age). When tumor sizes reached approximately 200–300 mm^3,
   animals were randomized into vehicle and treatment groups (8 mice
   each). For the treatment of tumor-bearing mice, erlotinib was
   administered at a dose of 5 mg/kg/day via gavage. Tumor volumes were
   recorded by blind measurements with a caliper and calculated as
   length × width^2 × 0.52. The Institutional Animal Care and Use
   Committee (IACUC) of Ren Ji Hospital approved all animal protocols.

Statistical analysis

   The sequencing data were deposited in the NCBI BioProject database
   under the accession number PRJNA973187. TCGA data were downloaded from
   the UCSC Xena Explorer and processed as previously described [[191]58].
   The Kaplan–Meier method with log-rank test was used for survival
   analysis. Pathway enrichment analysis of differentially expressed genes
   was performed using the clusterProfiler R package (v3.18.0) and
   Metascape ([192]http://metascape.org) [[193]59]. Gene set enrichment
   analysis was performed using the GSEA software (v4.1.0). In all
   experiments, comparisons between two groups were based on two-sided
   Student’s t-tests. Pearson’s correlation coefficient was used to
   measure the linear correlation between two variables. Graphics and
   statistics were generated using GraphPad Prism (v8.0) or R (v4.1.0).
   P-values of <0.05 were considered statistically significant.

Supplementary information

   [194]supplementary Figure^ (15.2MB, pdf)
   [195]Supplementary Tables^ (25KB, xlsx)
   [196]original data^ (20.9MB, pdf)

Author contributions

   GZ, XZ, LW, ZY, and YF designed and conceived the project. SZ, JZ, KY,
   JL, PM, and QZ performed the experiments. JW, CD, HL, and HJ collected
   clinical samples. MCC, JZ, TC, and YS analyzed the data and organized
   the figures. SZ and JZ wrote the manuscript. LT, GZ, and YF supervised
   the work and performed writing–review and editing. All authors verified
   the data and approved the final version of the manuscript.

Funding

   This work was supported by the National Natural Science Foundation of
   China (82172596 and 82373351 to GZ; 82273273 to YF; 82072560 to XZ),
   Shanghai Municipal Health Commission (202140049 to MCC), Shanghai
   Science and Technology Plan Project Shanghai Sailing Program
   (23YF1441100 to HL), Science and Technology Commission of Shanghai
   Municipality (23ZR1438800 to PM; 21Y11913900 to YF), Shanghai Municipal
   Education Commission-Gaofeng Clinical Medicine Grant Support (20161313
   to GZ), Collaborative Innovation Center for Clinical and Translational
   Science by Ministry of Education & Shanghai (CCTS-2022203 to GZ),
   innovative research team of high-level local universities in Shanghai
   (SHSMU-ZLCX20210200 to GZ), and 111 project (no. [197]B21024 to GZ).

Data availability

   Original images of immunoblotting are provided in the Supplementary
   information. All data and materials generated in this study are
   available upon request from the corresponding authors.

Competing interests

   The authors declare no competing interests.

Footnotes

   Edited by Roberto Mantovani

   Publisher’s note Springer Nature remains neutral with regard to
   jurisdictional claims in published maps and institutional affiliations.

   These authors contributed equally: Shichao Zhou, Jingyu Zang, Mei-Chun
   Cai, Kaiyan Ye, Jin Liu.

Contributor Information

   Guanglei Zhuang, Email: zhuangguanglei@gmail.com.

   Xiaojing Zhao, Email: zhaoxiaojing@renji.com.

   Lan Wang, Email: wanglanzm@163.com.

   Yu Zhuang, Email: yuzhuang2002@163.com.

   Yujie Fu, Email: richardjiejie@hotmail.com.

Supplementary information

   The online version contains supplementary material available at
   10.1038/s41419-024-07239-8.

References