Abstract

   The testis-specific BET protein BRDT structurally resembles the
   ubiquitous BRD4 and is misexpressed in cancer, and we show that BRDT
   misexpression may affect lung cancer progression. BRDT knockdown in
   lung cancer cells slowed tumor growth and prolonged survival in a
   xenograft model. Comparative characterization of PTEFb complex
   participation and chromatin binding indicates BRD4-redundant and
   BRD4-distinct BRDT functions. Unlike dual depletion, individual BRD4 or
   BRDT knockdown did not impair transcriptional responses to hypoxia in
   BRDT-expressing cells, consistent with redundant function. However,
   BRD4 depletion/BRDT complementation revealed that BRDT can also release
   paused RNA polymerase II independently of its bromodomains as we
   previously demonstrated not to be required for Pol II pause/release
   function of BRD4, underscoring the functional importance of the
   C-terminal domains in both BRD4 and BRDT and their potential as
   therapeutic targets in solid tumors. Based on this study, future
   investigations should explore BRD4-distinct BRDT functions and BRDT
   misexpression driving cancer pathogenesis.
     __________________________________________________________________

   BRDT, a testis-specific BET protein, is ectopically expressed in lung
   cancer, promoting hypoxia response and tumor progression.

INTRODUCTION

   Transcriptional regulation is essential for appropriate cellular
   responses to variable growth conditions and other external stimuli. In
   metazoans, which require precise spatiotemporal regulation of
   transcription for proper development, the physical release of RNA
   polymerase II (Pol II) from a promoter-proximal pausing elongation
   control “checkpoint” is a key regulatory step for the transcription of
   thousands of genes ([46]1, [47]2). Misregulation of this pause-release
   step, which occurs due to mutations that interfere with or alter the
   function of elongation factors, may drive or contribute to the
   pathogenesis of human cancers ([48]3). One prominent example is seen in
   mixed lineage leukemia (MLL), in which 11q23 translocation-driven
   oncofusions of the gene histone-lysine N-methyltransferase 2A (KMT2A)
   frequently occur with genes encoding components of the cyclin-dependent
   kinase 9 (CDK9)–containing super elongation complex (SEC) to create
   fusion proteins (e.g., MLL-AF9) that contribute to leukemogenesis by
   misdirecting SEC’s Pol II–releasing activity, hyperactivating
   transcription at MLL-controlled loci ([49]4–[50]6).

   The CDK9-containing elongation factor complex bromodomain-containing
   protein 4 (BRD4)–PTEFb (formed by BRD4 and the positive transcription
   elongation factor) is a master transcriptional regulator required for
   release of promoter-proximal paused Pol II into elongation at the
   majority of genes under normal conditions in somatic cells
   ([51]7–[52]12). However, recent studies have also demonstrated an
   essential role for the testis-specific BRD4 paralog bromodomain
   testis-specific protein (BRDT; in concert with the transcription factor
   A-MYB) in Pol II pausing and elongation control during meiosis in
   spermatogenic germ cells ([53]13–[54]15). BRDT was first identified as
   a cancer-testis antigen in lung cancers and however was later found to
   be ectopically expressed in other various cancer types ([55]16,
   [56]17). Ectopically expressed BRDT has been implicated as a
   transcriptional regulator for a set of genes driving cell proliferation
   in ovarian cancer and for a distinct set of genes driving cell
   migration in esophageal cancer, suggesting cancer context-specific
   roles ([57]18, [58]19). However, the functional role and impact of
   ectopic BRDT expression in lung cancer remain to be elucidated.

   Similar to BRD4, BRDT is a member of the BET (bromodomain and
   extraterminal domain) family of transcription factors, which also
   includes BRD2 and BRD3. Each BET protein has two tandem bromodomains
   (BD), which bind to acetylated histone lysines with high affinity, and
   an extraterminal domain (ET), which is thought to mediate transcription
   factor recruitment. However, only BRD4 and BRDT have the C-terminal
   domain that, at least for BRD4, is known to be required for interaction
   with the PTEFb subunits CDK9 and cyclin T1 (CCNT1) ([59]20, [60]21).
   Despite the structural similarities between BRD4 and BRDT, it is
   unclear whether the functions of ectopically expressed BRDT are unique
   and/or redundant to the functions of the ubiquitously expressed BRD4 in
   cancer cells.

   Here, we use in vitro and in vivo models to identify ectopically
   expressed BRDT as a factor promoting tumor growth, indicating BRDT as a
   potential therapeutic target in lung cancers with ectopic BRDT
   expression. Mechanistically, we show that ectopically expressed BRDT
   forms an active PTEFb complex that colocalizes with active
   transcription in lung cancer cells; that it can compensate for BRD4
   loss in the transcriptional response to hypoxic stress; and that,
   similar to BRD4, it can release paused Pol II independently of its
   bromodomains. Based on these findings, we propose that targeting the
   C-terminal PTEFb-interacting domain of BRDT could be a more effective
   therapeutic strategy than bromodomain targeting in cancers where BRDT
   is ectopically expressed.

RESULTS

BRDT is ectopically misexpressed in human lung cancers

   Because BRDT has been presumed to function similarly to BRD4, we
   initially investigated a potential compensatory role for BRDT in
   cancer. We analyzed the expression level of BRDT compared to BRD4
   across different cancer cell lines (n = 1479) and found no correlation
   between expression of BRDT and BRD4. However, we observed that more
   than 40% (n = 19) of the lines with robust BRDT expression (TPM ≥1,
   n = 45) were derived from samples of patients with lung cancer
   ([61]Fig. 1A). To further explore the role of BRDT in lung cancer, we
   then analyzed BRDT expression in patient samples within The Cancer
   Genome Atlas (TCGA) database. BRDT expression levels were significantly
   decreased in testicular germ cell tumors (TGCTs) relative to matched
   normal tissue controls ([62]Fig. 1B). However, BRDT is also ectopically
   expressed in several non-testis cancer types, most notably in lung
   adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), which
   are two major types of non–small cell lung cancer ([63]Fig. 1B).
   Together, LUAD and LUSC account for more than 65% of all lung cancers
   ([64]22).

Fig. 1. BRDT is ectopically expressed in lung cancers.

   [65]Fig. 1.
   [66]Open in a new tab

   (A) Plot showing the transcript expression levels of the ubiquitously
   expressed BRD4 versus the testis-specific BRDT across different cancer
   cell lines. Lung cancer–derived cell lines are labeled. Expression
   levels are presented as log[2]-transformed TPM (transcripts per
   million) with a pseudocount of 1 [log[2] (TPM + 1)]. (Expression data
   downloaded from DepMap.). (B) Elevated gene expression of BRDT in
   various solid tumors (T, in red) compared to matched normal tissue
   samples (N, in black) from the TCGA project database (analyzed via the
   GEPIA2 web server). BRCA (breast-invasive carcinoma), ESCA (esophageal
   carcinoma), HNSC (head and neck squamous cell carcinoma), LUAD, LUSC,
   STAD (stomach adenocarcinoma), TGCT, and UCS (uterine carcinosarcoma).
   Matched normal samples were obtained from TCGA and Genotype-Tissue
   Expression Program (GTEx). Expression levels are presented as log[2]
   (TPM + 1). (C) Kaplan-Meier curves comparing overall survival in
   patients with defined subtypes of LUAD (proximal inflammatory, proximal
   proliferative, and terminal respiratory unit) or LUSC (basal
   expression, classical expression, primitive expression, and secretory
   expression) whose tumors express low (below median, blue) versus high
   (above median, red) levels of BRDT (left) or BRD4 (right). The 95%
   confidence intervals are plotted as dashed lines. The y axis represents
   the percentage of patients surviving (percent survival) at a given time
   interval after diagnosis, represented on the x axis (months). HR was
   calculated on the basis of the Cox proportional hazards model, taking
   the high expression group as the intervention group [HR (high)].

   Next, we used TCGA data to perform comparative analysis of overall
   survival for patients with high versus low BRDT expression. Across
   eight different cancer types including LUAD and LUSC, there was no
   clear association between BRDT expression level and hazard ratio (HR)
   except for testicular germ cell cancer (TGCT), where BRDT is
   endogenously expressed (fig. S1A). In TGCT, higher BRDT expression
   appeared to be associated with increased HR (≫1), whereas higher BRD4
   expression appeared to be associated with decreased HR (≪1) (fig. S1A).
   The Kaplan-Meier analysis of overall survival in all patients with LUAD
   or LUSC (including both defined and undefined subtypes) revealed no
   clear impact of BRDT (or BRD4) expression on overall survival (fig.
   S1B). When limiting the analysis to patients with defined LUAD and LUSC
   subtypes, potential associations between BRDT expression and overall
   survival became apparent but were not statistically significant
   ([67]Fig. 1C). However, the impacts of high BRDT expression on overall
   survival were notable and statistically significant (HR = 6.3,
   P = 0.0063; log rank P = 0.0022) for the secretory subtype of LUSC
   (fig. S1C). Together, these data led us to hypothesize that ectopic
   BRDT expression might contribute to tumor growth or progression in
   certain forms of lung cancer.

Dox-induced BRDT knockdown impairs tumor growth in vitro and in vivo

   To evaluate our hypothesis regarding a potential role for BRDT in tumor
   growth or progression, we first generated a polyclonal antibody against
   BRDT (fig. S2A) and used it to confirm that ectopic BRDT expression is
   robustly detectable in cell lines derived from patients with small cell
   lung cancer, LUAD, and LUSC (fig. S2B). Next, to directly interrogate a
   role for BRDT in supporting tumor growth, we generated several cell
   lines with doxycycline (Dox)–inducible short hairpin RNAs (shRNAs) in
   the LUAD-derived NCI-H2009 background, which has been reported to
   readily generate xenograft tumors in nude mice ([68]23). Although our
   homemade BRDT antibody has several nonspecific bands, using dBET6
   treatment (which degrades BRDT) as a positive control confirmed
   specific knockdown of BRDT upon Dox treatment in these shBRDT lines
   ([69]Fig. 2A). Following these validation steps, we tested the impact
   of BRDT knockdown on colony formation and observed a notable reduction
   in the number and size of colonies formed in the presence of Dox,
   whereas shCtrl had no effect, excluding possible confounding effects of
   Dox treatment ([70]Fig. 2B). We then investigated the role of BRDT in
   tumor growth in vivo by picking a representative shBRDT line,
   reconfirming knockdown efficiency, and injecting cells subcutaneously
   into nude mice to generate xenograft tumors ([71]Fig. 2C). Compared to
   the untreated control group, tumor growth was significantly impaired,
   and the overall survival was significantly extended in the Dox-treated
   BRDT knockdown group ([72]Fig. 2, C and D). Last, we performed
   immunohistochemistry (IHC) on tissue slices from the excised xenograft
   tumors, first staining for BRD4 and BRDT to confirm specific knockdown
   of BRDT. Further IHC staining revealed lower levels of the
   proliferation marker Ki-67 and higher levels of the apoptosis marker
   TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine
   triphosphate nick end labeling) upon BRDT knockdown in these tumors
   ([73]Fig. 2E). In addition to indicating that loss of ectopic BRDT
   expression due to shRNA knockdown in NCI-H2009 cells impairs
   xenograft-derived tumor growth, these data suggest proproliferative
   and/or anti-apoptotic functions for BRDT when it is ectopically
   expressed in cancer.

Fig. 2. Dox-induced BRDT knockdown impairs tumor growth in vitro and in vivo.

   [74]Fig. 2.
   [75]Open in a new tab

   (A) Western blot analysis of Dox-induced BRDT knockdown in LUAD-derived
   NCI-H2009 cells with stable integration of BRDT-targeting (#1 to #3) or
   nontargeting (Scr) TetO-shRNA cassettes. dBET6 treatment in parental
   cells served as a positive control for BRDT depletion. (B) Colony
   formation assay in NCI-H2009 cells with integrated TetO-shRNA cassettes
   treated with or without Dox. (C) Left: Western blot confirming
   Dox-induced shRNA knockdown of BRDT expression in NCI-H2009 TetO-shBRDT
   cells before subcutaneous implantation into nude mice. Middle:
   Normalized tumor volume over the course of post-implantation treatment
   with phosphate-buffered saline (PBS) (n = 8, black) or Dox (n = 9,
   red). Right: Normalized tumor volume at day 44 post-implantation. BRDT
   knockdown significantly decreased NCI-H2009 xenograft tumor volume
   (**P = 0.0069, unpaired t test). (D) Left: Representative images of
   mice and excised tumors following treatment with PBS or Dox. Right:
   Kaplan-Meier curves comparing overall post-implantation survival. BRDT
   knockdown significantly increased overall survival in mice with
   NCI-H2009 xenograft tumors (****P < 0.0001, log rank test). (E)
   Representative histopathological staining of tumor tissue from mice
   treated with PBS (left) or Dox (right). Hematoxylin and eosin (H&E)
   staining is shown at top. IHC (with hematoxylin counterstain) is shown
   for the proliferation marker Ki-67, the apoptosis marker TUNEL, BRDT,
   and BRD4, with quantification of IHC-positive cells in four
   high-powered fields in each tumor (mean and SD shown). Scale bars, 50
   μm (for BRDs) and 20 μm (others). Statistical analysis was performed
   using the unpaired t test. BRDT knockdown significantly reduced %
   Ki-67^+ cells (****P < 0.0001) and significantly increased % TUNEL^+
   cells (****P < 0.0001). Significant reduction in % BRDT^+ cells
   (**P = 0.0015) confirmed successful BRDT knockdown in vivo. No change
   was observed in % BRD4^+ cells.

BRDT and BRD4 form both similar and distinct complexes with PTEFb on active
chromatin

   BRDT shares structural similarities including a putative
   PTEFb-interacting C-terminal motif with its related BET family member
   BRD4, which is a master regulator of Pol II transcription (fig. S2A).
   Despite these structural similarities, the function of ectopically
   expressed BRDT and its impact on Pol II transcription remain unclear.
   To comparatively characterize BRDT-PTEFb and BRD4-PTEFb complexes, we
   first conducted size exclusion chromatography using whole nuclear
   extracts from NCI-H510 lung cancer patient-derived cells, which express
   relatively high levels of BRDT at both the mRNA and protein level
   ([76]Fig. 1A and fig. S2B). We observed mutually exclusive interactions
   of BRDT and BRD4 with the PTEFb subunits CCNT1 and CDK9 in distinct
   complexes, suggesting that BRDT-PTEFb and BRD4-PTEFb may have distinct
   and independent functions ([77]Fig. 3A). To further characterize
   BRDT-PTEFb versus BRD4-PTEFb function at the chromatin level, we then
   performed chromatin immunoprecipitation sequencing (ChIP-seq) in
   NCI-H510 cells. We observed widespread colocalization of BRDT, BRD4,
   and CDK9 at active chromatin regions ([78]Fig. 3, B and C), with the
   vast majority of BRDT and CDK9 peaks overlapping BRD4 peaks ([79]Fig.
   3D). Pathway enrichment analysis (fig. S2C) using total peaks from BRD4
   and BRDT ChIP-seq revealed both shared and distinct pathways regulated
   by BRD4 and/or BRDT. Genome feature distribution analysis (fig. S2D)
   revealed a bias toward promoter-proximal BRDT peaks relative to a
   broader distribution of BRD4 peaks among different genomic features,
   consistent with BRD4 function at enhancers. Together, these results
   suggest that ectopic BRDT expression in tumors might (i) confer
   BRDT-specific functions and (ii) compensate for or augment some
   redundant subset of BRD4 functions.

Fig. 3. BRDT and BRD4 form similar and distinct complexes with PTEFb on
active chromatin.

   [80]Fig. 3.
   [81]Open in a new tab

   (A) Western blot for BRDT, BRD4, CCNT1, and CDK9 after size exclusion
   chromatography performed using benzonase-treated whole nuclear extracts
   from NCI-H510 cells. (B) Representative track example of ChIP-seq
   signal in NCI-H510 cells visualized with zoom-out view, showing
   colocalization of BRDT, BRD4, Pol II, and CDK9 at active chromatin
   regions marked by H3K27ac. (C) Global heatmaps of ChIP-seq signals for
   BRD4, CDK9, BRDT, and Pol II indicating their genome-wide
   colocalization at active chromatin regions (marked by H3K4me1, H3K4me3,
   and H3K27ac and lack of H3K27me3). (D) Venn diagram for peak overlaps
   among BRD4, BRDT, and CDK9 genome wide.

BRDT can compensate for loss of BRD4 in the transcriptional response to
hypoxic stress

   It has been well-demonstrated that BRD4 serves as a master regulator
   for global release of promoter-proximal paused Pol II and that it does
   so by recruiting PTEFb, which in turns phosphorylates the C-terminal
   domain of Pol II. Either BRD4 depletion or CDK9 inhibition leads to Pol
   II pausing at promoters, genome wide ([82]8, [83]10, [84]12). We have
   also demonstrated that when cells encounter the heat shock stress
   condition, BRD4 is sidelined and SEC acts as the essential
   PTEFb-recruiting mediator of Pol II release at heat shock–responsive
   genes ([85]10). However, this is not the case when cells encounter
   hypoxia (low oxygen tension): We recently identified BRD4 but not SEC
   as required for the gene-specific transcriptional response to hypoxic
   stress ([86]24). Hypoxia is particularly common in the core of solid
   tumors ([87]25, [88]26). Therefore, to investigate potential functional
   redundancy between BRD4 and BRDT in BRDT-expressing cancer cells, we
   decided to assess the impact of BRD4 versus BRDT loss on the
   transcriptional response to hypoxic stress. First, we generated
   separate BRD4 and BRDT knockdown lines in the NCI-H510 background using
   CRISPR-Cas9 and two individual single-guide RNAs (sgRNAs) for either
   factor ([89]Fig. 4A). We also used the BET-specific degrader dBET6 to
   rapidly deplete both BRD4 and BRDT in a subset of sgCtrl cells. After
   six hours of culture in normoxic (~18% O[2]) or hypoxic (1% O[2])
   conditions, we harvested sgCtrl, sgCtrl + dBET6, sgBRD4, and sgBRDT
   cells and divided the cell pellet into two portions, one for Western
   blot to validate knockdown efficiency and another for RNA sequencing
   (RNA-seq) to assess the impact of knockdown of these two factors on the
   transcriptional response to hypoxic stress ([90]Fig. 4A). Depletion of
   bands corresponding to BRD4 and BRDT indicated that knockdown via sgRNA
   was successful ([91]Fig. 4B). We then performed RNA-seq, which
   confirmed successful knockdown (fig. S3A) and captured hypoxia-induced
   gene expression changes, including hundreds of genes up-regulated after
   6 hours of hypoxia, as well as a smaller number of down-regulated genes
   ([92]Fig. 4C). Principal components analysis (PCA) analysis of RNA-seq
   data confirmed that cells in conditions of hypoxia or normoxia were
   transcriptionally distinct and that cells treated with dBET6 (dual
   BRD4/BRDT depletion) were transcriptionally distinct from cells with
   individual knockdown of either BRD4 or BRDT ([93]Fig. 4D). Gene
   Ontology (GO) analysis also confirmed that the top pathways enriched
   among the hypoxia-induced transcripts in untreated sgCtrl cells were
   all related to hypoxic response (fig. S3B). Clustering of the
   hypoxia-responsive genes in sgCtrl cells revealed a large subset of
   genes for which transcriptional induction failed upon dBET6 treatment
   but not upon individual knockdown of BRD4 or BRDT, indicating
   functional redundancy between these two factors in the transcriptional
   response to hypoxia ([94]Fig. 4, E and F). Last, GO term analysis
   revealed that the top pathways enriched among this redundantly
   BRD4/BRDT-regulated subset were all related to angiogenesis (fig. S3C).

Fig. 4. BRDT can compensate for loss of BRD4 in the transcriptional response
to hypoxic stress.

   [95]Fig. 4.
   [96]Open in a new tab

   (A) Schematic diagram of the experimental design for evaluating the
   impact of BRD4, BRDT, or dual BRD4/BRDT loss on transcriptional
   responses to hypoxia. (B) Western blot for BRD4, BRDT, and CDK9 in
   NCI-H510 cells under conditions of normoxia (~18% oxygen, left) or
   hypoxia (1% oxygen, right) after transfection of sgRNA targeting BRD4
   or BRDT, showing the efficiency of BRD4 or BRDT knockdown by each
   sgRNA. Nontargeting sgRNA (sgCtrl) is the negative control, and dBET6
   treatment (dual depletion of both BRD4 and BRDT) is the positive
   control for knockdown efficiency at the protein level. (C) Volcano plot
   of genes differentially expressed in response to hypoxic stress in the
   cells described in (B), with representative hypoxia-induced genes
   highlighted. Neither BRD4 nor BRDT was affected by hypoxia. (D) PCA
   analysis of RNA-seq data in cells cultured under normoxic or hypoxic
   conditions after depletion of either BRD4 or BRDT via individual sgRNA
   or after dual depletion of both BRD4 and BRDT by dBET6 treatment [as
   described in (B)]. (E) Track visualization of RNA-seq signal at the
   representative hypoxia-responsive gene NDRG1 locus showing
   transcriptional activity induced by hypoxia that was lost upon
   simultaneous depletion of BRD4 and BRDT by dBET6 treatment but remained
   upon individual depletion of either BRD4 or BRDT by sgRNA. (F) Heatmap
   showing a cluster of hypoxia-responsive genes for which transcriptional
   induction in response to hypoxia was strongly diminished by dBET6
   treatment, but not by sgRNA targeting either BRD4 or BRDT. PC,
   principal component.

BRDT releases promoter-proximal paused Pol II in a manner similar to that of
BRD4

   We have recently demonstrated that BRD4 mediates the release of
   promoter-proximal paused Pol II in a bromodomain-independent but
   C-terminal domain–dependent manner ([97]27). In addition to sharing
   structural similarities with BRD4, including a PTEFb-interacting
   C-terminal domain that is absent in BRD2 and BRD3 (Fig. S2A), BRDT also
   shares a redundant function in the transcriptional response to hypoxia
   ([98]Fig. 4 and fig. S3). We therefore hypothesized that BRDT might
   mediate the release of promoter-proximal paused Pol II in a manner
   similar to that of BRD4. To test this hypothesis, we first generated
   lentivirus-packaged, Dox-inducible, and green fluorescent protein
   (GFP)–tagged constructs of full-length BRDT and a bromodomain-less
   C-terminal BRDT fragment ([99]Fig. 5A). The C-terminal BRDT fragment
   was the same length as the C-terminal BRD4 fragment that we previously
   demonstrated to be sufficient to release Pol II ([100]27). Next, we
   used our BRD4-IAA7 degron system in the BRDT-null DLD-1 line, which
   enables rapid BRD4 depletion and Pol II pausing upon auxin treatment
   ([101]27). After stably integrating the GFP-tagged BRDT constructs in
   the BRD4-IAA7 DLD-1 background, we induced their expression and carried
   out GFP immunoprecipitation (IP) to confirm that BRDT interacts with
   PTEFb and that this interaction is retained in the absence of BRDT
   bromodomains ([102]Fig. 5B). After this validation step, we carried out
   a rescue experiment using our full-length BRD4 construct ([103]27) as a
   control for rescue of pause-release (as assessed by Pol II ChIP-seq)
   after the depletion of endogenous BRD4 ([104]Fig. 5, C to F). We
   confirmed the expression of the GFP-tagged BRD4 and BRDT constructs by
   Western blot using BRD4- or BRDT-specific antibodies (fig. S4A).
   Genomic integration and Dox-induced expression of the BRD4 and BRDT
   constructs were also confirmed by Pol II ChIP-seq reads that mapped to
   exons at the respective gene loci (fig. S4B). In the rescue experiment,
   BRD4 depletion caused accumulation of promoter-proximal paused Pol II
   ([105]Fig. 5C), reflected in genome-wide Pol II pausing ([106]Fig. 5D)
   that can be quantified as a pause-release ratio (log[2]PRR) of Pol II
   signal at gene bodies relative to promoters ([107]Fig. 5, E and F). Pol
   II pausing was rescued as expected by the expression of the full-length
   BRD4 construct, and it was similarly rescued by the expression of
   either the full-length or the bromodomain-less C-terminal fragment BRDT
   constructs ([108]Fig. 5, C to F). Together, these findings indicate
   that the bromodomain-independent function of BRD4 in the release of
   promoter-proximal paused Pol II is conserved in BRDT.

Fig. 5. BRDT releases promoter-proximal paused Pol II in a manner similar to
that of BRD4.

   [109]Fig. 5.
   [110]Open in a new tab

   (A) Schematic diagram of the Dox-inducible, N-terminally GFP-tagged
   BRDT constructs BRDT-FL (full-length BRDT) and BRDT-C (bromodomain-less
   C-terminal BRDT fragment). (B) Western blot for GFP and the PTEFb
   subunits CCNT1 and CDK9 after immunoprecipitation of the GFP-tagged
   BRDT constructs showing that both BRDT-FL and BRDT-C can interact with
   PTEFb. (C) Representative track example at the HSPA8 gene locus of Pol
   II ChIP-seq showing Pol II pausing upon depletion of endogenous BRD4 by
   auxin treatment and comparable rescue by BRD4-FL, BRDT-FL, and BRDT-C.
   (D) Heatmap of log[2]FC in Pol II occupancy for the rescue experiment
   in (C) showing a genome-wide increase in Pol II signal at promoters and
   a decrease in signal at gene bodies upon auxin-induced BRD4 depletion.
   BRD4-FL, BRDT-FL, and BRDT-C all rescue the genome-wide shift in Pol II
   occupancy in a similar manner. (E) Boxplot of log[2]PRR calculated from
   Pol II ChIP-seq signal at promoters versus gene bodies for the rescue
   experiment in (C). (F) Empirical cumulative distribution function plot
   of log[2]PRR for the rescue experiment in (C). (G) Putative model for
   ectopically expressed BRDT in lung cancer promoting tumor progression.
   Testis-specific BRDT has also been found in multiple cancers, with
   especially high expression levels found in common forms of lung cancer.
   BRDT knockdown (KD) impairs colony formation for lung cancer cells in
   vitro and reduces xenograft tumor growth in vivo, extending animal
   survival. BRDT can mediate gene expression by partnering with PTEFb to
   release paused Pol II, similar to BRD4, and BRDT can also functionally
   compensate for BRD4 in the transcriptional response to hypoxia. The C
   terminus of BRDT could be a potential therapeutic target for solid
   tumors, where hypoxic gene expression is critical.

DISCUSSION

   In this study, we have investigated the impact and function of the
   testis-specific transcription factor BRDT when it is ectopically
   expressed in lung cancer. We focused on ectopic expression of BRDT in
   lung cancer, the context in which BRDT was first identified as a
   cancer-testis antigen, based on the trends we identified in our
   analysis of publicly available expression data ([111]Fig. 1 and fig.
   S1). We observed consistently higher BRDT expression in cell lines
   derived from lung tumors relative to other lineages and highest levels
   of ectopic BDRT expression in lung tumors relative to other tumor
   types, but it is not clear whether BRDT expression in patient samples
   is entirely attributable to cancer cells or whether it might also
   reflect expression in tumor-associated immune cells. It is interesting
   to note that high expression of BRDT is associated with an increased HR
   in TGCTs, where BRDT expression is significantly lower than in normal
   tissue controls. However, while the lung cancer types LUAD and LUSC
   both ectopically express BRDT, high BRDT expression appears to have
   opposite impacts on HR in LUAD versus LUSC. Shedding some light on this
   apparently differential impact of ectopic BRDT expression, further
   analysis of BRDT isoforms revealed distinct patterns of isoform
   representation in these lung cancer types, including alternative
   full-length BRDT isoforms predicted to have an altered or absent first
   bromodomain primarily expressed in LUSC but not LUAD (fig. S4, D and
   E). Future studies will investigate the potential impacts and functions
   of BRDT isoforms in various forms of lung and other BRDT-expressing
   cancers.

   We have demonstrated that loss of ectopic BRDT expression via
   Dox-inducible shRNA knockdown in lung cancer cells inhibited colony
   formation and growth in vitro, an effect that was recapitulated in vivo
   by decreased tumor size and prolonged survival in a xenograft model
   ([112]Fig. 2). Although the impacts of BRDT loss on tumor size and
   survival were statistically significant, they were not particularly
   notable or complete. Given the apparent redundancy in BDRT and BRD4
   function evidenced by our ChIP-seq studies ([113]Fig. 3), the loss of
   ectopically expressed BRDT may be compensated for by BRD4.
   Alternatively, shRNA knockdown may not affect all BRDT isoforms.

   Based on the structural similarity between BRDT and BRD4 and after
   demonstrating BRDT interaction with PTEFb and similar binding for BRDT
   and BRD4 genome wide ([114]Fig. 3), we have examined the long-proposed
   yet untested redundancy between these bromodomain-containing factors.
   We did so by focusing on the transcriptional response to the solid
   tumor-relevant stress condition of hypoxia, in which we recently
   demonstrated that BRD4 plays a major role. In addition to demonstrating
   that BRDT can compensate for the loss of BRD4 in the transcriptional
   response to hypoxia ([115]Fig. 4), we have also demonstrated that BRDT
   can mediate the release of promoter-proximal paused Pol II
   independently of its tandem bromodomains ([116]Fig. 5), similar to the
   distinct “layer” of BRD4-PTEFb we recently identified as actively
   promoting release of paused Pol II without binding to histone lysine
   acetylation ([117]27). Based on lessons learned from BET inhibitors
   ([118]28–[119]31), we would predict limited antitumor efficacy for
   inhibitors that target the BET bromodomains of ectopically expressed
   BRDT. Instead, the C-terminal region of BRDT should be considered as a
   therapeutic target if the goal is to interfere with BRDT-mediated
   release of paused Pol II.

   Our findings do not indicate that Pol II release is either the sole or
   primary function of BRDT when it is ectopically expressed in cancer
   cells. Functional redundancy between BRDT and BRD4 is also context and
   gene specific. For example, our data indicate that expression of the
   transcription factor Myc is controlled more by BRD4 than BRDT (fig.
   S4C). Moreover, although BRD4 is also expressed in germ cells, it
   cannot compensate for loss of BRDT as a meiotic factor during
   spermatogenesis; both JQ1 treatment and BRDT depletion impair
   spermatogenesis, indicating that unlike its function as a Pol
   II–releasing factor, BRDT’s function as a meiotic factor is likely to
   be bromodomain dependent ([120]21, [121]32, [122]33). The disparate
   nature of the bromodomains in BRDT and BRD4 could be an important
   difference contributing to the distinct, nonredundant functions of
   these BET proteins. Studies have shown that unlike BRD4, in which both
   bromodomains recognize acetylated histone tails, in BRDT, only the
   first bromodomain (BD1) binds with high affinity to histone lysine
   acetylation, while the second bromodomain (BD2) contributes a
   nonspecific interaction with DNA that is essential for BRDT function in
   meiotic and post-meiotic chromatin remodeling and genome reorganization
   ([123]21, [124]34). Further studies are required to determine the
   functions of BRDT bromodomains when BRDT is ectopically expressed in
   somatic cells. Bromodomains are altered or absent in some of the BRDT
   isoforms ectopically expressed in cancer (fig. S4, D and E). Short BRD4
   isoforms (lacking the C terminus PTEFb-interacting domain but retaining
   the bromodomains and ET domain) have been well studied, with functions
   demonstrated to be different from or opposite to that of the
   full-length BRD4 isoform ([125]35–[126]39). Although BRDT and BRD4
   expression levels are uncorrelated in lung cancer cell lines, the
   potential relationship between expression of full-length versus short
   BRD4 and BRDT isoforms is unclear. Distinct, nonredundant functions of
   BRDT and BRD4 could also be conferred by the formation of distinct
   complexes with PTEFb, as implicated by our size exclusion study
   ([127]Fig. 3). We are currently working to elucidate the components and
   the functional significance of the two distinct BRDT-PTEFb complexes we
   have identified in BRDT-expressing lung cancer cells.

   Last, it will be interesting to investigate the epigenetic mechanisms
   by which BRDT, normally restricted to the testis, comes to be
   ectopically expressed in neoplasia arising from somatic cells. We know
   that BRDT expression gradually increases during spermatogenesis, with
   detectable mRNA appearing at the early spermatocyte stage, but upstream
   regulators are unknown ([128]21). Do the same mechanisms functioning in
   spermatogenesis turn on BRDT expression in lung and other
   BRDT-expressing cancers? To identify upstream regulators of BRDT,
   activation would not only elucidate the molecular mechanisms by which
   testis-specific transcription factors such as BRDT are ectopically
   expressed in cancer but might also reveal mechanisms of dysfunction in
   spermatogenesis. CRISPR screening is a reasonable route to achieve this
   goal.

MATERIALS AND METHODS

Cell culture

   NCI-H510 (HTB-184), NCI-H1385 (CRL-5867), NCI-H838 (CRL-5844),
   NCI-H2009 (CRL-5911), and DLD-1 (CCL-221) cells were purchased from
   American Type Culture Collection. DLD-1 cells were cultured in
   Dulbecco’s modified Eagle’s medium (Corning, #10013CV), while other
   cell lines were in RPMI (Thermo Fisher Scientific, #A1049101). Both
   media were supplemented with 10% fetal bovine serum (Sigma-Aldrich,
   #F2442), 1% Penicillin-Streptomycin (Gibco, #15140122), and 1% GlutMAX
   (Gibco, #35050061). Cells were cultured in an incubator at 37°C with a
   supply of 5% CO[2]. Hypoxia induction was carried out in a Whitley H35
   Hypoxystation with 1% O[2].

Western blot

   Following procedures previously described ([129]27), for whole-cell
   lysate Western, cells were directly lysed in 2X Laemmli sample buffer
   (Bio-Rad, #1610747) and boiled for 10 min before loading to the precast
   protein gels (Bio-Rad, #4561096). For GFP IP, cells were trypsinized
   and washed with phosphate-buffered saline (PBS) twice. Cell pellets
   were incubated with cold Triton X-100 lysis buffer [50 mM tris HCl (pH
   8.0), 150 mM NaCl, 1.5 mM MgCl[2], 10% glycerol, 0.5% Triton X-100,
   dithiothreitol, benzonase, protease inhibitor, and phosphatase
   inhibitor] and rotated at 4°C for ≥1 hour. EDTA (final concentration of
   1.5 mM) was added to the cell lysate before centrifugation for 20 min
   at max speed at 4°C. Supernatant was collected and incubated with
   GFP-Trap magnetic beads (Proteintech, #gtma) at 4°C with rotation for
   ≥4 hours. After immunoprecipitation, beads were washed with cold Triton
   X-100 lysis buffer five times at 4°C and eluted with 2.5 M glycine
   buffer (pH 2.0) at room temperature for 5 min. A 1 M tris base pH > 10
   was added to neutralize the eluate before equal volume of 4X Laemmli
   sample buffer was added. Samples were then boiled for 5 min for Western
   blot. BRD4 (#13440), BRD2 (#5848), β-actin (#3700), CDK9 (#2316), BRD3
   (#94032), and cyclin T1 (#81464) antibodies were purchased from Cell
   Signaling Technology. GFP antibody (#sc-9996) was purchased from Santa
   Cruz Biotechnology. BRDT antibody was generated in house.

Knockdown by shRNAs or sgRNAs

   TetO-shRNA constructs were cloned using Age I and Eco RI sites in the
   Tet-pLKO-puro vector (#21915) obtained from Addgene. shRNA target
   sequences are as follows: TetO-shScr (CCTAAGGTTAAGTCGCCCTCGCTCGA),
   TetO-shBRDT#1 (CCGATGGATCTTGGAACTATT), TetO-shBRDT#2
   (CCCTAAGTTTACAGAAGTAAA), and TetO-shBRDT#3 (CATCAGAAGCTCAAGATAAAT).

   sgRNA constructs were cloned using Bsm BI v2 sites in the lentiCRISPR
   v2 vector (#52961) obtained from Addgene. The sgRNA target sequence
   information was obtained from the Addgene Brunello library. Target
   sequences are as follows: sgCtrl (GCACTACCAGAGCTAACTCA), sgBRD4#1
   (CACCAAACTCCTGAGCATCA), sgBRD4#2 (CCAGACCCCTGTCATGACAG), sgBRDT#1
   (AACTCCCTGGAGATAAACTT), and sgBRDT#2 (CAACTCCAGTTCACAAACTG).

   Like previously described ([130]27), all target sequences in the
   constructs were verified by Sanger sequencing (ACGT, Inc.) before
   lentiviral packaging in 293T cells with pspax2/pmd2.g plasmids.
   Transfection was carried out using the Lipofectamine 3000 Transfection
   Reagent (Invitrogen, #L3000001) in a six-well plate. Lentivirus was
   collected and filtered before infection. Infected cells were selected
   using Puromycin (Gibco, #A1113803) until no further cell death
   observed. To induce the shRNAs, 500 nM Dox (STEMCELL Technologies,
   #72742) was added to the cell culture and replaced every 2 days.

BRD4 degron cell line and BRD4/BRDT mutant expressing lines

   BRD4-IAA7 degron cells expressing Dox-inducible GFP and GFP-BRD4-FL
   were generated in previous study ([131]27). Similarly, constructs for
   expressing Dox-inducible GFP-BRDT-FL and GFP-BRDT-C were cloned by
   using NEBuilder HiFi DNA assembly reaction [New England Biolabs (NEB),
   #E2621S] with synthetic gBlocks (Integrated DNA Technologies and Twist
   Biosciences). Lentivirus was packaged to infect the BRD4-IAA7 cells,
   and blasticidin (Gibco, #A1113903) was used to select transduced cells
   for 2 weeks. Mutant expression was achieved by adding 50 nM Dox to the
   cell culture for 2 days, and endogenous BRD4 depletion was achieved by
   treating the cells with Auxin (Abcam, #ab146403) for 3 hours.

Colony formation assay

   A total of 400 cells were seeded in each of the six-well plate. Dox
   treatment was started the next day and replaced together with fresh
   media every 2 days for 18 days. A 4% paraformaldehyde solution (Thermo
   Fisher Scientific, #50-980-495) in PBS was used to fix the cells at
   room temperature under shaking for 20 min. Plates were then washed
   three times gently with tap water and dried overnight, followed by
   violet solution (MilliporeSigma, #HT90132) staining for 20 min.

Xenograft studies

   Six-week-old female athymic mice (nu/nu genotype, BALB/c background)
   were purchased from Envigo and housed under aseptic conditions.
   NCI-H2009 TetO-shBRDT#1 cells were implanted into the flank of athymic
   mice as previously described ([132]40). Briefly, 4 × 10^6 cells, in
   0.4 ml of cell culture media with Matrigel (BD Bioscience) at 1:1
   ratio, were injected in the right flank of mice under anesthetization
   by inhaled isoflurane. Mice were randomly assigned to vehicle (no Dox,
   n = 8) and Dox treatment [20 mg/kg by intraperitoneal injection for 14
   days followed by dox water (2 mg/ml) for 14 days, n = 9] groups when
   the size of tumor reached at 100 mm^3 (18 days after implantation). The
   tumor sizes were measured on alternate days with calipers, and the mice
   were euthanized when the tumor size reached 1000 mm^3. Dox used in the
   animal studies was purchased from Sigma-Aldrich (catalog no.
   D9891-25G). Subcutaneous tumors were collected from the mouse and fixed
   with 4% paraformaldehyde (PFA). PFA-fixed tumors were paraffin-embedded
   and sectioned (4 μm) for hematoxylin and eosin staining. For IHC
   staining, anti-Ki67 (2 μg/ml) (Ventana, Tucson, AZ, USA), BRD4 (1:1000;
   #13440, Cell Signaling Technology), and the home-made BRDT (1:1000;
   rabbit) antibodies were used. To assay the apoptotic response to
   treatment, TUNEL staining was performed using the DeadEnd Colorimetric
   TUNEL system (Promega, Madison, WI, USA) according to the
   manufacturer’s protocol. All animal protocols were approved by the
   Northwestern University Institutional Animal Care and Use Committee
   (#IS00000787).

Chromatin immunoprecipitation sequencing

   ChIP-seq experiments for DLD-1 cells were carried out as previously
   described ([133]27). Briefly, cells in a 15-cm plate were crosslinked
   by 15 ml of 1% PFA (Thermo Fisher Scientific, #28908) in PBS with
   shaking for 10 min at room temperature. A 2 ml of 2.5 M gylcine was
   added to quench the reaction with continue shaking for 5 min.
   Crosslinked cells were then scraped and collected. Covaris E200 was
   used for chromatin sonication with following settings: 10% duty factor,
   200 cycles per burst, and 140 peak intensity power for 10 min. For
   NCI-H510 cells, 5 μl of Rpb1 (Cell Signaling Technology, #14958), CDK9
   (Cell Signaling Technology, #2316), H3K27ac (Cell Signaling Technology,
   #8173), BRD4 (Bethyl Laboratories, #A700-004), or 40 μl of BRDT
   anti-sera were incubated with the lysate overnight at 4°C with rotation
   for immunoprecipitation. Protein G–coupled Dynabeads (Invitrogen,
   #10004D) were used to pull down the immune complexes. Protease K
   (Roche, #3115828001) was used during reverse crosslink (65°C, 1200 rpm,
   overnight). DNA was then purified through QIAquick PCR Purification Kit
   (QIAGEN, #28106). Libraries were prepared by using the KAPA HTP library
   preparation kit (Roche, #07961901001) and sequenced on the Illumina
   NovaSeq 6000.

RNA sequencing

   NCI-H510 cells were cultured in rectangular flask (Corning, #430639).
   To deplete both BRD4 and BRDT, 250 nM dBET6 was added to the sgCtrl
   cells 3 hours before hypoxia induction. To carry out hypoxia induction,
   normoxia plates were kept in normal incubator, while hypoxia plates
   were moved to the hypoxia station. Six hours later, all plates were
   transferred to the cold room, and cell pellets were collected. The cell
   pellets (20%) were saved for Western blot and 80% for RNA extraction.
   Total RNAs were extracted using the RNeasy Mini Kit from QIAGEN
   (QIAGEN, #74106) according to the procedure. mRNAs were enriched using
   the NEBNext poly(A) mRNA magnetic isolation module (NEB, #E7490L).
   Libraries were prepared using the NEBNext Ultra II Directional RNA
   library preparation kit (NEB, #E7760L) and sequenced on the NovaSeq
   6000 ([134]27).

Data analysis

   The survival analysis based on the expression of BRDT and BRD4 genes
   was performed using data from TCGA implemented in GEPIA2 portal
   ([135]41). BRDT isoforms were annotated using genome assembly
   GRCh38.p13 and Ensembl database, release 98 ([136]42). Survival was
   calculated as the overall survival in months since the time of
   diagnosis, and the median gene expression was used for the high/low
   expression group cutoff and log rank test for hypothesis testing. HR
   calculation was based on Cox PH model with the 95% confidence interval.
   GEPIA2 was used to generate Kaplan-Meier plots.

Chromatin immunoprecipitation sequencing

   ChIP-seq reads in the NCI-H510 cells were at least 30 M per sample and
   were aligned to hg19.

   ChIP-seq reads in DLD-1 were aligned to hg38 with bowtie ½ ([137]43).
   Following procedures previously described ([138]27), A total of 6481
   Pol II–transcribed genes with transcription start site (TSS) and
   transcription end site (TES) annotated were obtained from our
   previously published study ([139]44). Heatmap for the log[2] fold
   change (log[2]FC) of Pol II occupancy was generated using the deeptools
   with nondefault options: binSize of 10 and pseudocount of 0.1. PRR was
   calculated as the ratio of Pol II reads density, which was computed by
   FeatureCounts 2.0 at the genebody [300 base pair (bp) downstream of the
   TSS to the TES) to the promoter (100 bp upstream to 300 bp downstream
   of the TSS). Empirical cumulative distribution functions and boxplots
   for the log[2]PRR were generated in R. Tracks were visualized in igv
   2.13.2 (Broad Institute).

RNA sequencing

   Preprocessing was performed using CETO Toolbox ([140]45). Sequence
   quality was assessed using FastQC v0.11.2 ([141]46), and quality
   trimming was done using Trimmomatic ([142]47) with parameters
   TRAILING:30 MINLEN:20. RNA-seq reads were aligned to the GRCh38 genome
   using STAR v.2.5.2 ([143]48), and only uniquely mapped reads with a
   two-mismatch threshold were considered for downstream analysis and
   quantified to the gene level using HTSeq ([144]49). Output BAM files
   were converted into bigWig track files to display coverage throughout
   the genome (in rpm). Gene count tables were constructed following
   Ensembl GRCh38.p13, release 98 ([145]42). Raw read counts were
   normalized to remove technical differences and batch effects using R
   library RUVseq 1.26.0 ([146]50), specifically the upper-quartile
   correction. Differential gene expression analysis was performed using
   DESeq2 1.32.0 ([147]49). Genes with adjusted P values <0.05 (the
   Benjamini and Hochberg method) and an absolute log[2]FC ≥1 were
   considered differentially expressed.

Acknowledgments