MYC regulates ribosome biogenesis and mitochondrial gene expression programs through interaction with Host Cell Factor-1

The oncoprotein transcription factor MYC is a major driver of malignancy and a highly-validated but challenging target for development of anti-cancer therapies. Novel strategies to inhibit MYC may come from understanding the co-factors it uses to drive pro-tumorigenic gene expression programs, providing their role in MYC activity is understood. Here, we interrogate how one MYC co-factor, Host Cell Factor (HCF)-1, contributes to MYC activity in a Burkitt lymphoma setting. We identify genes connected to mitochondrial function and ribosome biogenesis as direct MYC/HCF-1 targets, and demonstrate how modulation of the MYC–HCF-1 interaction influences cell growth, metabolite profiles, global gene expression patterns, and tumor growth in vivo. This work defines HCF-1 as a critical MYC co-factor, places the MYC–HCF-1 interaction in biological context, and highlights HCF-1 as a focal point for development of novel anti-MYC therapies.


6
To monitor the contribution of the MYC-HCF-1 interaction to cell proliferation, we pulsed each of our 166 engineered Ramos lines with 4-OHT for two hours to generate approximately equally mixed populations 167 of switched and unswitched cells. We then monitored how the GFP-positive switched cells in the 168 population compared to their unswitched counterparts in terms of glutamine-dependency ( Figure 1F), cell 169 cycle profiles ( Figure 1G), and long-term growth ( Figure 1H). We see that 4A switched cells have a 170 selective advantage over the WT switch in their ability to grow without exogenous glutamine ( Figure 1F).

171
This advantage is likely due to loss of the MYC-HCF-1 interaction, as the VP16 HBM mutant cells have 172 a corresponding deficit in growth under glutamine-starvation conditions ( Figure 1F). Compared to their 173 wild-type counterparts, cell cycle profiles for the two mutants are not dramatically altered, but we did 174 observe small but statistically significant changes in the proportion of cells in G2/M ( Figure 1G), which 175 again trend in opposite directions for the two MYC mutants-decreasing for the 4A-expressing cells and 176 increasing for those that express the VP16 HBM mutant ( Figure 1G). And finally, by monitoring long-term 177 growth, we observe that 4A mutant cells are gradually lost from the culture over time, whereas there is a 178 significant enrichment of VP16 HBM cells, compared to the WT control switch ( Figure 1H). The altered 179 and opposing impact of the 4A and VP16 HBM mutations in these assays leads us to conclude that the 180 MYC-HCF-1 interaction promotes the glutamine-dependency-and rapid proliferative status-of these 181 BL cells in culture.

183
The MYC-HCF-1 interaction influences intracellular amino acid levels changes in intracellular amino acid levels are not confined to aspartic acid and glutamine, but rather there 7 is a general tendency for amino acid levels to be increased in 4A and decreased in VP16 HBM mutant 211 cells, compared to the WT switch (Table 1). Based on these data, we conclude that the MYC-HCF-1 212 interaction, directly or indirectly, plays a global role in influencing intracellular amino acid levels in this 213 setting.

215
The MYC-HCF-1 interaction influences expression of genes connected to ribosome biogenesis

231
If anti-correlations between these gain-and loss-of-function mutants can be used to reveal MYC-HCF-1 232 co-regulated processes, the above data highlight protein synthesis and mitochondrial function as key   Figure 3E). This analysis confirms that reciprocal changes we observed for the GO 243 categories in Figure 3B and 3C results from reciprocal changes in the expression of a common set of 244 genes. From our data, we conclude that the MYC-HCF-1 interaction plays an important role in influencing 245 the expression of genes that promote ribosome biogenesis and maintain mitochondrial function.

247
Finally, we interrogated our RNA-Seq dataset for transcript changes that would correlate with the 248 widespread changes in amino acid levels that occur upon modulation of the MYC-HCF-1 interaction.

249
Here, we discovered that the accumulation of amino acids we observe with the 4A mutant is generally

258
Ribosome biogenesis and mitochondrial matrix genes respond rapidly to HCF-1 depletion 259 As a challenge to the concept that ribosome biogenesis and mitochondrial matrix genes are controlled 260 via the MYC-HCF-1 interaction, we asked whether expression of these genes is impacted by acute 261 depletion of HCF-1, mediated via the dTAG method (Nabet et al., 2018). We used CRISPR/Cas9-262 triggered homologous recombination to integrate an mCherry-P2A-FLAG-FKBP12 F36V cassette into the 263 HCFC1 locus in Ramos cells; the effect of which is to amino-terminally tag HCF-1N with the FLAG epitope

277
We performed RNA-Seq analysis three hours after addition of dTAG-47-a time point at which the 278 majority of HFC-1N is degraded ( Figure 4B). Despite the early timepoint, we identified ~4,500 significant  supplement 1G), and we identified a union set of ~450 genes-oppositely regulated by the 4A and VP16 291 HBM mutants-the expression of which also changes when HCF-1N is destroyed ( Figure 4F). Within this 292 set, loss of HCF-1N tends to mimic the loss of function 4A mutant ( Figure 4F), particularly for transcripts 293 that are reduced when the MYC-HCF-1 interaction is disrupted ( Figure 4G). Moreover, within the cohort 294 of transcripts that are reduced by both HCF-1N destruction and the 4A mutation, we see clear 308 peaks are enriched in DNA sequences linked to nuclear respiratory factor (NRF)-1, as well as the 309 Sp1/Sp2 family of transcription factors. Interestingly, although HCF-1 has not previously been linked 310 directly to NRF-1 or this element, the motif is also a functional, non-canonical, E-box that MYC proteins 311 are known to bind (Blackwell et al., 1993;Morrish et al., 2003). Consistent with enrichment of this variant 312 E-box element in the HCF-1N peaks, overlaying these data with our previous ChIP-Seq analysis of MYC

324
We previously reported that WDR5 has an important role in recruiting MYC to chromatin at a cohort of 325 genes overtly linked to protein synthesis, including more than half of the ribosomal protein genes (Thomas 326 et al., 2019). To determine whether these genes are also bound by HCF-1, we compared our HCF-1N 327 and MYC ChIP-Seq data to those we generated for WDR5 in this setting. Interestingly, there is little  10 Finally, we asked if recruitment of MYC or HCF-1 to chromatin at a subset of genes from Figure 5H

350
2A and 2B). Consistent with this result, we observe that the binding of MYC to chromatin in cells is not 351 significantly affected by these mutations ( Figure 5I). We also observe that binding of HCF-1 to these 352 same sites is insensitive to HBM mutations in MYC ( Figure 5J). The observation that binding of neither 353 protein to chromatin is HBM-sensitive strongly supports the idea MYC and HCF-1 interact to control the 354 expression of these genes through a co-recruitment-independent mechanism.

356
From these results, we conclude that, in this context, (i) MYC is a common binding partner with HCF-1

384
Next, we injected unswitched cells into the flanks of mice, allowed tumors to form, and then switched to 385 each of the MYC variants by injecting mice with tamoxifen ( Figure 6E). As we observed previously

405
Seq with that generated for the 4A MYC mutant in vitro ( Figure 3). Interestingly, more than 70% of the 406 genes repressed in the 4A cell line are also repressed in either 4A-1 or 4A-2 tumors, and there is a 407 common set of 942 genes that are shared between all three datasets ( Figure 6I). These genes coalesce 408 on those connected to ribosome biogenesis, translation, and the mitochondria ( Figure 6J). The overlap 409 of induced genes was less pronounced (30%; Figure 6K) and these genes are less clustered, although 410 we do observe modest enrichment in categories connected to metabolism, chromatin binding, and 411 transcription coregulator activity ( Figure 6L). The recurring connections we observe between the MYC-412 HCF-1 interaction and ribosome biogenesis and mitochondrial function, both in vitro and in vivo, strongly 413 supports the notion that a major function of this interaction is stimulate ribosome production and 414 mitochondrial vigor, and that these actions are central for the ability of MYC to drive tumor onset and 415 maintenance.

417
The wealth of MYC-interaction partners provides a rich resource for the discovery of novel ways to 418 eventually inhibit MYC in the clinic. Unfortunately, the complexity of the MYC interactome also presents 419 a barrier to prioritizing which co-factors to pursue. The highest priority co-factors should be those that 420 directly interact with MYC, play a critical role in the core tumorigenic functions of the protein, and where 421 there is proof-of-concept that disrupting interaction with MYC would provide a therapeutic benefit in the 422 context of an existing malignancy. Here, we provide this information for HCF-1. We show that MYC

430
Through the use of mutations that bidirectionally modulate the interaction between MYC and HCF-1,

440
Among the direct targets of the MYC−HCF-1 interaction are genes that catalyze rate-limiting steps in

451
The concept that there can be process-specific co-factors for MYC is not widely appreciated. Attention is

475
The concept that HCF-1 is a biomass-specific co-factor for MYC can also account for our discovery that

492
Our discovery that MYC and HCF-1 work together to regulate transcription via a co-recruitment-493 independent mechanism is surprising, but not without precedent. E2F transcription factors interact with

501
Finally, our demonstration that disrupting the MYC-HCF-1 interaction in the context of an existing tumor 502 promotes its regression provides compelling proof-of-concept for the idea that inhibitors of this interaction 503 could have utility as anti-cancer agents. Switching WT MYC to the 4A mutant caused rapid and 504 widespread induction of apoptosis, and was associated with changes in the expression of genes 505 connected to ribosome biogenesis, translation, and the mitochondria, consistent with the idea that 506 14 reduced expression of these MYC-HCF-1 target genes triggers the regression process. The small and 507 well-defined nature of the HBM suggests that, if structural information becomes available for the HCF-1 508 VIC domain, it could be possible to develop small molecule inhibitors that block the MYC-HCF-1 509 interaction. The most obvious concern with this strategy is that HCF-1 is not a MYC-specific co-factor, 510 and that its interactions with other transcription factors may limit or prevent attainment of a therapeutic 511 window. To our knowledge, MYC proteins and E2F3a are the only transcription factors that interact with 512 HCF-1 via an "imperfect" HBM, which we show here is sub-optimal for robust HCF-1 association. It might 513 be possible to develop a therapeutic window by exploiting the non-canonical nature of the HBM in MYC,

514
with the expectation that this interaction will be more sensitive than others that carry higher affinity HBM 515 motifs. We note, however, that many of the factors with which HCF-1 interacts via an HBM are inherently 516 pro-proliferative, with the E2F proteins in particular playing a predominant role in cancer initiation, 517 maintenance, and response to therapies (Kent and Leone, 2019). We also note that HCF-1 has been 518 reported to be overexpressed in cancer, and its overexpression can correlate with poor clinical outcomes 519 (Glinsky et al., 2005). It is possible, therefore, that on-target collateral consequences of inhibiting the 520 MYC-HCF-1 interaction could also have therapeutic benefit against malignancies.

713
The total number of sequencing reads for each replicate is shown in Supplementary File 3.

715
For qPCR, samples (either input or ChIP) were brought up to a final volume of 200 μl using TE. Each

813
Next generation sequencing analyses

857
Phenylalanine-D8 and Biotin-D2, were added to each sample to assess sample extraction quality.

858
Samples were subjected to protein precipitation by addition of 800 µl of ice cold methanol (4X by volume),

859
and incubated at -80°C overnight. Samples were centrifuged at 10,000 rpm for 10 minutes to eliminate

888
Full MS analyses were acquired over a mass range of m/z 70-1050 using electrospray ionization positive 889 mode. Full mass scan was used at a resolution of 120,000 with a scan rate of 3.5 Hz. The automatic gain 890 control (AGC) target was set at 1 × 10 6 ions, and maximum ion injection time was at 100 ms. Source

914
Tentative and putative identifications were determined within Progenesis QI using accurate mass 915 measurements (<5 ppm error), isotope distribution similarity, and fragmentation spectrum matching    Student's t-test between GFP-and GFP+ cells was used to calculate P-values; a = 0.033, b = 0.0041, c = 0.0006. (H) Switchable Ramos cells were pulsed with 4-OHT as in (F), and the proportion of GFP-positive cells measured by flow cytometry 24 hours after treatment and every three days following. For each of the replicates, the proportion of GFPpositive cells is normalized to that on day one. Shown are the mean and standard error for three biological replicates. Student's t-test between WT and each of the mutants at day 25 was used to calculate P-values; a=0.000028, b=0.00026. The following figure supplement is available for figure 1:    Figure 2J. (E) Metabolites in the "alanine, aspartate, and glutamate metabolism" pathway that are significantly (FDR < 0.05) impacted in the VP16 HBM mutant. Node color represents the fold-change over WT. The remainder of pathway is shown in Figure 2J.      and correlative (bottom) in direction between the 4A and VP16 HBM mutants. (E) and (F) Categories from the top eight families in GO enrichment analysis of the correlative gene clusters shown in (C), for genes that were decreased (E) or increased (F) with both mutants. The Q-value of categories is represented by the ribbon color, which is scaled across these figure and Figures 3D and 3E. Categories are ranked by the number of matched genes, and genes are ranked by the number of categories into which they fall.   Data are smoothed with a cubic spline transformation. (G) Relationship between protein-coding genes that are co-bound by promoter-proximal MYC and HCF-1N by ChIP-Seq, and are significantly (FDR < 0.05) decreased or increased in response to the 4A or VP16 HBM mutations. Also shown are genes where the expression is anti-correlated between the 4A and VP16 HBM mutants. (H) Heatmap showing genes that are co-bound by promoter proximal MYC and HCF-1N in Ramos cells, have anti-correlative gene expression changes between for the 4A and VP16 MYC mutants, and have significant gene expression changes with HCF-1N degradation. Genes that fall into GO categories relating to ribosome biogenesis or translation (RiBi/translation), and mitochondrial function or metabolism (Mito/metabolism) are highlighted. (I) ChIP, using anti-HA antibody, was performed on parental or switchable Ramos cells treated for 24 hour with 20 nM 4-OHT. Enrichment of genomic DNA was monitored by qPCR using primers that amplify across peaks. HBB is a negative locus for HA-MYC. ChIP efficiency was measured based on the percent recovery from input DNA. Shown are the mean and standard error for three biological replicates. (J) ChIP, using anti-HCF-1N antibody, was performed on parental or switchable Ramos cells treated for 24 hour with 20 nM 4-OHT. Enrichment of genomic DNA was monitored by qPCR using primers that amplify across peaks. EIF4G3 and HBB are negative loci for HCF-1N. ChIP efficiency was measured based on the percent recovery from input DNA. Shown are the mean and standard error for three biological replicates.
The following source data and figure supplement(s) are available for figure 5:       Only days five to 19 are shown here; the full course of the experiment is depicted in Figure 6 -figure supplement 2A. Student's t-test between WT and each of the mutants was used to calculate P-value; *P < 0.000043. (C) Kaplan-Meier survival curves of mice (n=6 of each) injected with switched cells. Log-rank test was used to calculate P-value (< 0.0001) from six biological replicates. (D) PCR assays of genomic DNA was used to determine the proportion of switched cells present in each tumor after sacrifice. Each dot represents an individual tumor, and the line indicates the mean for each group. Student's t-test between WT and each of the mutants was used to calculate P-values; a = 0.0002, b < 0.0001. (E) Tumor maintenance schema: Unswitched WT, 4A-1, 4A-2, and ∆264 cells were injected into the flanks of nude mice. Tumors were grown until day 15, at which point mice received tamoxifen injections (one/day for three days) to induce switching of the cells. (F) Average tumor volume before and after cells were switched. The day at which tamoxifen (Tam) administration was initiated is indicated with an arrow. Shown are the mean and standard error for seven mice for WT and six mice for 4A-1, 4A-2, and ∆264 cells. (G) Kaplan-Meier survival curves of mice in the tumor maintenance assay (n=7 for WT, and n=6 for 4A-1, 4A-2, and ∆264).
The day at which tamoxifen (Tam) administration was initiated is indicated with an arrow. Log-rank test was used to calculate P-value (< 0.0001). (H) Annexin V staining and flow cytometry were performed on cells isolated from tumors at 48 and 96 hours following the first tamoxifen administration to determine the extent of apoptosis. Shown are the mean and standard error for four mice each. Student's t-test between WT and each of the mutants was used to calculate P-value; *P