Proteomic analysis identifies ZMYM 2 as endogenous binding partner of TBX 18 protein in 293 and A 549 cells

The TBX18 transcription factor regulates patterning and differentiation programs in the primordia of many organs yet the molecular complexes in which TBX18 resides to exert its crucial transcriptional function in these embryonic contexts have remained elusive. Here, we used 293 and A549 cells as an accessible cell source to search for endogenous protein interaction partners of TBX18 by an unbiased proteomic approach. We tagged endogenous TBX18 by CRISPR/Cas9 targeted genome editing with a triple FLAG peptide, and identified by anti-FLAG affinity purification and subsequent LC-MS analysis the ZMYM2 protein to be statistically enriched together with TBX18 in both 293 and A549 nuclear extracts. Using a variety of assays, we confirmed binding of TBX18 to ZMYM2, a component of the CoREST transcriptional corepressor complex. Tbx18 is coexpressed with Zmym2 in the mesenchymal compartment of the developing ureter of the mouse, and mutations in TBX18and in ZMYM2 were recently linked to congenital anomalies in the kidney and urinary tract (CAKUT) in line with a possible in vivo relevance of TBX18-ZMYM2 protein interaction in ureter development.


Introduction
T-box (Tbx) genes encode a large family of proteins that are characterized by a conserved DNA-binding domain, the T-box [1]. This 180 amino acid long region recognizes a short conserved stretch of DNA, the T-box binding element (TBE), which can occur in repeats of variable spacing and orientation [2,3]. Binding to these elements in the genome results in transcriptional changes of adjacent genes, the nature of which, i.e. activation or repression, depends on both the composition of the DNA binding site as well as on protein motifs outside the T-box domain to which transcriptional cofactors can bind [3][4][5][6]. Besides DNA-binding, the T-box domain mediates interaction with other transcription factors, which is likely to increase target gene specificity. Functional analysis characterized T-box transcription factors as essential regulators of various cellular processes in diverse tissue contexts during metazoan development [4,7]. Not surprisingly, mutations in T-box genes have been identified to cause a large set of human congenital diseases [8].

T-box 18 (TBX18) is a member of a vertebrate-specific subgroup of T-box genes.
Mutations in TBX18 were recently recognized to underlie congenital forms of urinary tract malformations in human patients [9]. These phenotypic changes relate to a function of TBX18 in the early development of the ureteric mesenchyme as shown by analysis of mice carrying a null allele of the gene [10,11]. Besides dilatations of the ureter and the renal pelvis (hydroureternephrosis) [10], mutant mice feature defects in the otic capsule and the otic fibrocyte compartment [12], in smooth muscles of the prostate [13], and in the pericardium and epicardium [14][15][16]. They also have a hypoplasia of the sinoatrial node and the caval vein myocardium [17,18], and malformation of the vertebral column and the rib cage, the latter of which causes postnatal lethality [19]. Defects in these organ systems were traced to changes in patterning and differentiation of the respective tissue primordia during early organogenesis [10][11][12]17,19]. To date, both the molecular complexes in which TBX18 acts as well as the transcriptional targets of its activity have remained unexplored in all of its expression domains.
Murine Tbx18 encodes a protein of 613 amino acid residues (aa) that can be divided in an N-terminal region (154 aa), a middle T-box domain (182 aa) and a large Cterminal region (277 aa) [20]. A couple of years ago, we started the biochemical characterization of the TBX18 protein. We found that the T-box domain of TBX18 Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20210642/927037/bcj-2021-0642.pdf by guest on 24 December 2021 preferentially binds to a palindromic set of two TBEs in vitro, and that the protein represses transcription upon binding to these elements in cellular transactivation assays. Transcriptional repression depends on a conserved EH1-motif in the Nterminal region that acts as a binding site for TLE3/GROUCHO corepressors [20]. GROUCHO proteins attenuate transcription by facilitating pausing of RNA polymerase II and/or by recruiting histone deacetylases to condensate chromatin [21][22][23][24].
The N-terminal region of TBX18 also harbors a short stretch of basic amino acid residues that is necessary and sufficient to mediate nuclear localization in cells [20].
No conserved protein motifs have been identified in the large C-terminal region.
However, TBX18 proteins lacking the C-terminal region lose part of their repression function indicating that additional cofactors for transcription modulation bind to this domain [9].
In vitro, TBX18 forms homodimers and heterodimerizes with related T-box proteins.
TBX18 also binds to members of the PAX, SIX and NKX homeobox families and to the Zn-finger transcription factor GATA4 [6,20,25,26]. In vivo relevance has been shown for the interaction with PAX3 in somite patterning and with SIX1 in ureter development [25,26].
More recently, we embarked on a proteomic approach to search for TBX18 protein interaction partners in an unbiased fashion. Using overexpressed dual tagged TBX18 as bait, we identified by tandem purification and subsequent liquid chromatography- However, overexpression in cells may force unnatural protein interactions, particularly when a relatively promiscuous protein interaction motif such as the T-box is involved. Therefore, we aimed to identify and characterize proteins with which TBX18 forms functionally relevant endogenous complexes in cells. Here, we present a proteomic screen for interaction partners of endogenously tagged TBX18 protein in 293 and A549 cells obtained through CRISPR/Cas9 gene editing. We identified the zinc finger protein ZMYM2 as a binding partner of TBX18 in both cell lines, validated its binding, provide data for its function as a transcriptional corepressor of TBX18 and discuss the relevance of TBX18 interaction with ZMYM2 in congenital anomalies of the kidney and the urinary tract (CAKUT).  [28]; for A549 cells, we used Lipofectamine 3000 (#L3000008, ThermoFisher). We plated 6x10E5 cells in one well of 6-well plates and transfected them with 2.5 µg of DNA and 10 µl of Lipofectamine diluted in 1000 µl of Opti-MEM (#31985062, ThermoFisher). The pd2EGFP-N1 plasmid (Clontech, Mountain View, CA, USA) was independently transfected to check for efficiency of transfection, which was verified by epifluorescence microscopy. Analysis of TBX18 mRNA expression in human cell lines was done by semiquantitative RT-PCR with normalization against GAPDH exactly as described before [27].

Generation of A549 and 293 cell lines with FLAG-tagged TBX18
We used a CRISPR/Cas9-mediated knock-in approach based on a protocol from the Medenhall and Myers laboratory [29] [29]. Surviving colonies were expanded and subsequently screened for correct integration by RFLP analysis.

Southern Blot
Genomic DNA from single cell derived cultures was screened for correct integration of the 3′-homology arm, the 5′-integration and neomycin by Southern blot analysis of  according to the manufacturer's instructions. Three independent repetitions for each cell line along with parental cell controls were run in one-well 10% SDS polyacrylamide gels, and the proteins were silver stained. The gels were delivered to the MHH Proteomics facility for fractionation and liquid chromatography-mass spectrometry (LC-MS) analysis. There, proteins were mixed, alkylated by acrylamide and further processed as described [34]. Peptide samples were analyzed with a shotgun approach and data dependent analysis in a LC-MS system (RSLC, LTQ Orbitrap

Identification of TBX18 interacting proteins in 293 and A549 cells by LC-MS
Velos, both Thermo Fisher) as described [34]. Raw MS data were processed using Proteome discoverer 1.4 (ThermoFisher) and Max Quant software (version 1.5) [35], and a database containing human and viral proteins and common contaminants. Annotation, Visualization and Integrated Discovery (DAVID) web program but no functional clustering from the annotations was found [36].

Expression constructs
An expression plasmid for ZMYM2. constructs for mutant forms of ZMYM2 from CAKUT patients were previously described in detail [37]. inverted microscope (Leica, Wetzlar, Germany).

Nuclear recruitment assays
To check for interactions inside the nucleus, we cotransfected 293 cells with plasmids encoding a tagged form of full-length ZMYM2 with a variant of TBX18 which lacked the nuclear localization signal (TBX18ΔNLS) [20]. Mutant forms of ZMYM2 from CAKUT patients were cotransfected with either full-length TBX18 or TBX18ΔNLS. One day after transfection, cells were fixed with PFA and immunofluorescent stainings against the different protein tags were carried out and documented with a DM6000 microscope. To check for physical interaction, we performed Glutathione S-Transferase (GST)

GST pull-down assays
fusion protein pull-downs as described previously [20]. GST Glycerol, 0.5% Triton X-100, 80 mM NaCl, pH 7.5). Analysis of the pulled-down proteins was achieved by separating the proteins by denaturing SDS-PAGE, heatdrying the resulting gels on filter paper and then exposing them to phosphorescent imaging plates for 1 day. Detection of the proteins was done on a FLA-7000 Laser photo documentation system (Fujifilm) with a sensitivity of S10000 and a pixel size of 50 µm. HiMark™ Prestained Protein Ladder, 10 to 250 kDa (#LC5699, ThermoFisher) was used for size determination.

Immunoprecipitation assays
To perform immunoprecipitations of endogenous proteins, we used whole cell

RNA in situ hybridization
10-m paraffin sections of the posterior trunk region of embryonic day (E) 12.5 wildtype NMRI mice were subjected to RNA hybridization with digoxigenin-labelled antisense riboprobes as previously described [38].
Since available antibodies proved unsuitable for precipitation of endogenous protein complexes of TBX18 in these cell lines in our hands, we decided to exploit the CRISPR/Cas9 technology to introduce a triple FLAG (3xFLAG) affinity tag fragment [40,41] into the TBX18 locus of both 293 and A549 cells [29,42]. Human TBX18 is

ZMYM2 interacts with endogenous TBX18 in the nucleus of 293 cells
To validate the finding of our proteomic approach, we independently assayed for endogenous interaction of ZMYM2 and TBX18 in 293 cells. Using a commercial antibody against ZMYM2 we detected a band running slightly higher than the expected size (154.9 kDa) in lysates of both 293 parental and 293-TBX18.3xFLAG knock-in cells ( Figure 3A). In immunofluorescent analysis, anti-FLAG or anti-ZMYM2 antibodies detected predominantly nuclear antigens in 293-TBX18.3xFLAG cells and a minor cytoplasmic localization of the respective proteins ( Figure 3B). To validate a possible direct interaction, we applied a proximity ligation assay. Using only a single antibody against TBX18.FLAG or ZMYM2 did not yield a staining. Combined application of both antibodies resulted in mainly nuclear PLA signals and little staining in the cytoplasm ( Figure 3C). Thus, we identified ZMYM2 as a mainly nuclear protein that interacts with TBX18 in 293-TBX18.3xFLAG cells.

Overexpressed ZMYM2 binds to TBX18 in the nucleus of 293 cells
To further validate ZMYM2-TBX18 interaction, we transiently overexpressed

ZMYM2.MYC-FLAG and TBX18.MYC proteins in 293 cells. After immunoprecipitation
with an anti-FLAG antibody, we detected both ZMYM2.MYC-FLAG and TBX18.MYC protein on a western blot by an anti-MYC antibody indicating coprecipitation ( Figure   4A). To further test whether overexpressed TBX18 productively interacts with ZMYM2 in 293 cells, we made use of a nuclear recruitment assay that we previously described [20]. It is based on the identification of a classical nuclear localization signal (NLS) at the N-terminus of the TBX18 protein. When this NLS is deleted, the resulting protein (TBX18ΔNLS) is excluded from the nucleus but can be shuttled back to this compartment by binding to a protein, which carries such a signal. Upon  Table S4A). Finally, we performed a PLA experiment with overexpressed proteins in 293 cells ( Figure 4D). Since TBX18 forms homodimers [20],

ZMYM2 interacts with the T-box region of TBX18
To delineate the region of the TBX18 protein, which interacts with ZMYM2, we performed pull-down assays using a previously described series of bacterially expressed fusion proteins of GST with the N-and C-terminal region and the T-box of TBX18 [20] and radioactively labeled ZMYM2 produced with an in vitro expression

system. ZMYM2 interacted with the T-box and with the N-terminal plus T-box region
of TBX18, indicating that the T-box of TBX18 serves as an interaction domain with ZMYM2 ( Figure 5).

ZMYM2 mutant proteins of CAKUT patients are partly compromised in binding to TBX18
Recent work identified mutations in ZMYM2 as a cause of congenital anomalies of the kidney and urinary tract (CAKUT) in human patients [37]. Since CAKUT have also been observed in patients with mutations in TBX18 [9], ZMYM2 and TBX18 may functionally interact in vivo.
To validate such a possibility, we analyzed whether Zmym2 is coexpressed with Tbx18 in the undifferentiated mesenchymal compartment of the embryonic murine ureter at a stage, when TBX18 has a critical function in this tissue [10,11]. In situ hybridization on transverse sections of the proximal ureter region of E12.5 embryos indeed found coexpression of the two genes in the ureteric mesenchyme. Expression of Zmym2 was also found in the inner epithelial tissue in agreement with previous reports that Zmym2 has a widespread low expression in the developing urogenital system [37] and many developing and mature organs (www.proteinatlas.com) [50]. We next asked whether mutations in ZMYM2 that have been characterized in CAKUT patients [37], compromise binding to TBX18. We therefore re-evaluated mutant ZMYM2 localization in 293 cells for the described truncated patients' variants p.Gly257*, p.Gln398*, p.Cys536; Leufs*13, p.Arg540*, p.Tyr763Glnfs*6, p.(Lys812Aspfs*18) and p.Cys823* by immunofluorescence against the MYC-tag (depicted in Figure 6B) and applied a nuclear recruitment assay with TBX18 ( Figure   6C). p.Tyr763* was the only truncated form that was predominantly expressed in the nucleus. Accordingly, TBX18ΔNLS.HA was used in the nuclear recruitment assay for this variant of ZMYM2. All other ZMYM2 variants, that were located almost completely in the cytosol, were tested with full length TBX18.HA protein for nuclear recruitment ( Figure 6C-E). Immunofluorescent images ( Figure 6C) and further statistical evaluation ( Figure 6D,E, Supplementary

ZMYM2 acts as a corepressor for TBX18
To analyze whether ZMYM2 modulates TBX18 transcriptional activity, we performed dual luciferase reporter assays in 293 cells. We used a recently described reporter plasmid in which a palindromic repeat of two TBEs is cloned in front of an SV40 minimal promoter and a firefly luciferase gene, together with expression constructs of  on TBX18 was abolished when both proteins were coexpressed ( Figure 7A). Of note, ZMYM2 had no effect on reporter expression in absence of TBX18 ( Figure 7B). ZMYM2 was shown to be a member of the CoREST corepressor complex, that harbors additional proteins such as KDM1A [45][46][47][48]. Since KDM1A was recovered in both of our proteomic screens in both 293 and A549 cells (Supplementary Table S2, S3), we tested it for its effect on TBX18 repressive activity. In fact, we found that KDM1A enhanced TBX18 repressive activity in a concentration-dependent manner.
Notably, the addition of ZMYM2 to KDM1A relieved the corepression exerted by KDM1A similar to the effect of ZMYM2 on TLE3/GROUCHO ( Figure 7C). Again, neither KDM1A alone nor the combination of KDM1A and ZMYM2 exerted transcriptional repressing activity on the luciferase reporter in absence of TBX18 protein ( Figure 7D). We conclude that ZMYM2 and KDM1A, members of the

Immunoprecipitation of an endogenous FLAG tagged TBX18 protein identifies largely distinct sets of interacting proteins in A549 and 293 cells
Our affinity mediated immunoprecipitation approach against an engineered FLAG tag at the C-terminus of TBX18 protein identified both in 293 and A549 cells a large number of proteins coprecipitating with TBX18. These extensive lists (115 and 270 proteins, respectively) comprised proteins localized outside the nucleus (e.g. cytoskeletal proteins) indicating that the nuclear extracts used for the immunoprecipitations were not purified to homogeneity. Furthermore, we detected many proteins (e.g. the ribonucleoproteins RALY and HNRNPN) that have been frequently detected in similar anti-FLAG immunoprecipitation experiments as documented in the crapome database [49]. They may represent proteins that are unspecifically recognized by the anti-FLAG antibody, that adhere strongly to the binding matrix or that may preferentially denature to form adhesive aggregates. In any case, our findings enforce the notion that affinity purification against a single tag provides more false positive candidates compared to a tandem purification strategy available for dual tagged proteins.
Given these considerations, we decided to increase the stringency of our approach by selecting proteins that were statistically significantly enriched in the three individual immunoprecipitations from 293-and A549-TBX18.3xFLAG cells, respectively. By this we ended up with 10 candidate proteins for 293-and 36 proteins for A549-TBX18.3xFLAG cells, respectively, of which only two were in common: TBX18, as expected, and the transcriptional corepressor ZMYM2. One may argue that the poor overlap reflects differential gene expression in the two cell lines. In fact, A549 and 293 cells derive from two different organs and exhibit cell-type specific Most notably, we found only few tissue-specific transcription factors, which all belonged to the Zn-finger family, and not to the homeobox protein family. It seems possible that in an overexpression approach, the endogenous binding partners of TBX18 are limited and that weak or even unnatural interactions are overrepresented or newly established. Such a behavior is eased by the fact that the T-box seems to act rather promiscuously by binding to many very different types of proteins or motifs.
Alternatively, TBX18 expression in A549 and 293 cells may be extremely low to interact with very few other proteins. Irrespective of the precise mechanism, it shows that the expression level (e.g. low endogenous vs highly overexpressed) of a protein and the mode of purification (single tag vs dual tag) can dramatically affect the identification of relevant binding partners in cellular systems.

ZMYM2 may present a biologically meaningful protein interaction partner of TBX18
ZMYM2 was the protein with the highest significance score found in our proteomics analyses both from 293 and A549 immunoprecipitates. ZMYM2 interacted with endogenous and overexpressed TBX18 in the nucleus of 293 cells, and bound to the T-box of TBX18 in vitro. Moreover, ZMYM2 enhanced the repressive activity of TBX18 in cellular transactivation assays but did not affect reporter activity on its own, providing a strong hint that this Zn-finger protein acts a corepressor for TBX18 in our experimental system.
Such a notion is supported by previous findings that ZMYM2 is a nuclear protein that binds to chromatin and recruits a repressive complex consisting of the histone demethylase KDM1A/LSD1, the cofactor CoREST (RCOR1 or RCOR2), and a histone deacetylase (HDAC1 or HDAC2) [48,51]. We recently showed that RCOR3,

Data Availability
All supporting data are included within the main article and its Supplementary Files.

Competing Interests
The authors declare that there are no competing interests associated with the manuscript.      p<0.05 and ***, p<0.001 as determined in Student's t-test. Note that protein expression can be found both in the nuclear and cytosolic compartment. Therefore, addition of nuclear and cytosolic localization can add to more than 100% in D and E.
(F) Summary table of the TBX18/ZMYM2 nuclear recruitment assay.