How a Hidden Molecular Feedback Loop Helps Colorectal Cancer Hide From the Immune System

In an interview with Pharmacy Times, Ajay Goel, PhD, AGAF, chair of City of Hope's Department of Molecular Diagnostics and Experimental Therapeutics, and Junyong Weng, PhD, a City of Hope visiting scholar, discussed their team’s newly published research uncovering a self-reinforcing regulatory circuit between the RNA acetyltransferase NAT10 and the oncogene MYC in colorectal cancer. Their data were presented at the American Association for Cancer Research Annual Meeting 2026.

The study identifies how this bidirectional loop drives autophagy-mediated degradation of MHC-I molecules, enabling tumor cells to evade immune detection—a finding with particular relevance for patients with microsatellite-stable (MSS) colorectal cancer (CRC), a subgroup that derives little benefit from existing checkpoint immunotherapy.

Goel and Weng explored the mechanistic evidence underlying the NAT10–MYC axis, the models used to validate immune evasion, and the translational potential of NAT10 inhibition as a strategy to restore antigen presentation and sensitize immune-cold tumors to checkpoint blockade.

Pharmacy Times: Can you walk us through the feedback loop between NAT10 and MYC — specifically, how does each component reinforce the other, and what is the evidence that this is a true regulatory loop rather than a unidirectional relationship?

Ajay Goel, Ph.D., AGAF: In our study, we identified a bidirectional positive feedback loop between NAT10 and MYC. On one side, we found that MYC directly activates NAT10 transcription. When we depleted MYC, the NAT10 protein expression decreased, and global ac4C levels also dropped. We then performed promoter analysis and identified multiple putative MYC-binding sites in the NAT10 promoter, including canonical E-box motifs. Importantly, ChIP-qPCR confirmed direct MYC occupancy at these NAT10 promoter regions.

On the other side, we found that NAT10 reinforces MYC at the posttranscriptional level. MYC was identified among NAT10-bound and ac4C-modified transcripts. NAT10 knockout reduced MYC mRNA and protein levels and accelerated MYC mRNA decay. In addition, NAT10-RIP and ac4C-RIP confirmed direct binding of NAT10 to MYC mRNA and showed that NAT10 stabilizes MYC through ac4C modification.

The key point is that our data supports reciprocal regulation in both directions: MYC transcriptionally activates NAT10, while NAT10 stabilizes MYC mRNA. That is why we interpret this as a true regulatory loop rather than a simple one-way relationship. At the same time, if we want to be maximally rigorous in discussion, we can acknowledge that a full rescue/epistasis series would make the loop even stronger mechanistically. Still, the current evidence already supports a bona fide MYC–NAT10–MYC circuit.

Pharmacy Times: How does autophagy serve as the mechanistic bridge between NAT10-MYC activity and MHC-I degradation? Is this canonical or selective autophagy, and how was the specificity for MHC-I established experimentally?

Junyong Weng, PhD: In our model, autophagy is the key mechanistic bridge linking the NAT10–MYC axis to immune escape. We found that NAT10 promotes the stability of multiple autophagy-related transcripts through ac4C modification, including BECN1, ATG3, ATG5, and VPS33A. This leads to increased autophagic activity, which in turn promotes lysosomal degradation of MHC-I, thereby reducing antigen presentation on tumor cells.
We supported this at several levels.

First, KEGG and DAVID analyses showed that NAT10-correlated genes and NAT10-bound/ac4C-modified transcripts were enriched in autophagy, mitophagy, lysosome, and related pathways. Second, NAT10 loss reduced key autophagy transcripts and proteins, decreased LC3B-II, reduced LC3 puncta, and impaired autophagic flux in both BafA1-based and reporter-based assays. Third, when we examined MHC-I degradation directly, we found that BafA1 caused clear accumulation of MHC-I. In contrast, the proteasome inhibitor MG132 did not restore it, indicating that MHC-I was predominantly degraded via the autophagy–lysosome pathway rather than the proteasome.

We also showed that NAT10 knockout increased MHC-I expression in CRC cells and in vivo tumors, and confocal microscopy demonstrated strong colocalization between MHC-I and LC3B puncta, supporting autophagic targeting of MHC-I. Functionally, NAT10-low tumors also showed higher HLA-A/B/C and greater CD8 infiltration in clinical tissues.

Regarding whether this is canonical or selective autophagy, our data most strongly support autophagy-dependent lysosomal degradation of MHC-I. We clearly show the pathway dependence, but we do not yet fully define the cargo-recognition machinery in the strict sense of classical receptor-mediated selective autophagy. The careful wording would be that we demonstrate autophagy–lysosome–dependent MHC-I degradation, while the precise selectivity machinery remains to be further resolved.

Pharmacy Times: What models—cell lines, organoids, or in vivo systems—were used to validate immune evasion, and how well do they recapitulate the immunosuppressive tumor microenvironment seen in human colorectal cancer?

Goel: In our study, we used a multilayered set of models. For mechanistic studies, we used the human CRC cell lines RKO and SW620. For immune-related and in vivo studies, we used the murine CRC cell lines CT26 and MC38. We generated NAT10-knockout models in these systems and then used them in both in vitro and in vivo assays.

For immune functional studies, we used OT-I splenic T-cell coculture with OVA-expressing MC38 cells to test antigen-specific T-cell killing. In vivo, we used subcutaneous syngeneic tumor models in immunocompetent BALB/c and C57BL/6J mice, with anti–PD-1 treatment and CD8 depletion experiments. These models allowed us to evaluate tumor growth, checkpoint response, CD8 dependence, T-cell infiltration, and cytokine production under immunocompetent conditions.

We also integrated extensive human data, including TCGA, GEO, FUSCC, TMDU, IHC, and multiplex immunofluorescence. These human data sets were important because they showed that high NAT10 is associated with MSS status, lower immune and ESTIMATE scores, lower IPS, reduced HLA-A/B/C expression, and lower CD8 infiltration, which closely matches the immune-cold phenotype we model experimentally.

That said, we should also be honest about the limitations. Our in vivo work relies mainly on subcutaneous syngeneic models, which are highly useful for immunotherapy studies but do not fully recapitulate the complexity of the human colorectal tumor microenvironment, particularly the location-specific stromal, vascular, and microbial influences. I would say our models capture the core immune-evasion biology quite well, but there is still room for future validation in orthotopic or organoid-based systems.

Pharmacy Times: Is NAT10 overexpression or MYC amplification correlated with specific colorectal cancer subtypes, and does that have implications for which patients might benefit most from targeting this axis?

Weng: Our findings suggest that this axis is particularly relevant in MSS/proficient DNA mismatch repair (pMMR) colorectal cancer, the subgroup that derives very limited benefit from checkpoint blockade. We found that NAT10 expression is significantly higher in MSS/pMMR tumors than in MSI-H/dMMR tumors across TCGA, our validation cohorts, and IHC analyses. We also found that NAT10-high tumors display reduced immune infiltration and poorer predicted response to immunotherapy.
More specifically, our study points to CMS2 MSS-CRC as a particularly relevant subtype. CMS2 is characterized by MYC hyperactivation, reduced MHC-I expression, and an immune-desert phenotype, and our data show coordinated upregulation of MYC and NAT10, together with reduced HLA-A/B/C, in this context. That places the NAT10–MYC loop in a very specific biological and clinical setting.

From a translational standpoint, the patients most likely to benefit from targeting this axis would probably be those with MSS/pMMR, MYC-high, NAT10-high, HLA-low, immune-cold tumors, especially within the CMS2-like context. One nuance I would keep in mind is that, in our study, the strongest evidence is for MYC activation/expression rather than MYC amplification alone. When answering live, I would frame it as a MYC-activated subtype rather than limiting it to amplification.

Pharmacy Times: Given that MSS colorectal cancer is largely resistant to immune checkpoint blockade, how does NAT10-mediated MHC-I loss contribute to or compound this resistance, and could restoring MHC-I expression be sufficient to sensitize these tumors?

Goel: In our view, NAT10-mediated MHC-I loss contributes to MSS resistance by making tumor cells even less visible to cytotoxic T cells. A cold microenvironment already characterizes MSS CRC, but our data suggest that NAT10 further compounds this by promoting autophagy-mediated degradation of MHC-I, thereby weakening antigen presentation and limiting effective CD8 T-cell responses.

Several layers of evidence support this. NAT10-high tumors had lower HLA-related scores, lower MHC-I expression, reduced T-cell activation signatures, lower IPS, and poorer predicted response to checkpoint therapy. Conversely, NAT10 loss increased MHC-I, improved CD8 infiltration, enhanced IFN-γ and TNF production, strengthened T-cell killing, and markedly sensitized tumors to anti–PD-1 treatment. In some mice, the combination even produced complete tumor regression.

As for whether restoring MHC-I alone is sufficient, I would answer carefully: Our findings indicate that restoring MHC-I is a major and likely necessary part of the sensitization process, but probably not the entire story. In the CD8 depletion experiments, the antitumor effect of NAT10 loss was largely reversed, indicating that CD8 immunity is central. However, a residual difference remained even after CD8 depletion, suggesting that additional immune mechanisms may also contribute.

So I would say that MHC-I restoration is highly important and likely one of the dominant mechanisms, but it is not necessarily sufficient on its own in every context.

Pharmacy Times: NAT10 is an ac4C RNA acetyltransferase with broad targets across the transcriptome. How do the authors rule out that the immune evasion phenotype is driven by off-target mRNA acetylation changes rather than the MYC-specific loop they describe?

Weng: I would answer this directly and honestly: In our study, we do not claim that the phenotype is exclusively MYC-dependent. NAT10 is indeed a broad epitranscriptomic regulator, and our data show that it stabilizes multiple autophagy-related transcripts in addition to MYC. So our model is not that NAT10 acts only through 1 downstream target. Rather, our interpretation is that MYC is a central organizing node within a broader NAT10-driven program.
What makes the MYC loop especially compelling is that it is mechanistically supported in both directions: MYC directly activates NAT10 transcription, and NAT10 directly stabilizes MYC mRNA through ac4C. That gives the axis a self-reinforcing property that can amplify the broader autophagy and immune-evasion program. At the same time, our data also show direct stabilization of autophagy regulators, including BECN1, ATG3, ATG5, and VPS33A. So the phenotype likely reflects both the MYC loop and direct regulation of autophagy machinery by NAT10.

The most accurate answer is that we do not fully exclude broader transcriptome-wide contributions. Instead, our findings support the idea that the MYC–NAT10 loop is a core driver embedded within a larger NAT10-dependent epitranscriptomic network. If we wanted to push this even further mechanistically, rescue experiments would help define how much of the phenotype is specifically MYC-dependent.

Pharmacy Times: What therapeutic strategies are proposed or could be envisioned to disrupt the NAT10-MYC loop? Are there existing autophagy inhibitors, NAT10 inhibitors (eg, Remodelin), or MYC-targeting approaches that could be combined with immunotherapy, and what are the key translational obstacles?

Goel: The most directly supported strategy from our study is pharmacologic NAT10 inhibition, particularly with Remodelin, combined with PD-1 blockade. In our in vivo CT26 model, Remodelin alone significantly inhibited tumor growth, and the combination of Remodelin with anti–PD-1 further suppressed tumor progression beyond either monotherapy. Remodelin also increased T-cell infiltration, which supports the idea that NAT10 inhibition can remodel the immune microenvironment and improve checkpoint response.

Conceptually, there are several therapeutic angles here. One is direct NAT10 inhibition, which we think is attractive because NAT10 is more druggable than MYC and may be more tumor-selective than broadly suppressing autophagy. Another is combining NAT10 inhibition with immunotherapy, especially in immune-cold MSS disease. A third possibility is combining with autophagy/lysosome-targeting approaches. However, our study also emphasizes that global autophagy inhibition is limited by toxicity because autophagy is essential for normal tissue homeostasis. MYC-targeting strategies are conceptually relevant upstream, but the main reason we focused on NAT10 is precisely that MYC remains very difficult to target directly.

The key translational obstacles are, first, selectivity and safety, because NAT10 regulates many transcripts across the transcriptome. Second, biomarker selection will be critical; we likely need to identify the right subset, such as MSS/CMS2/NAT10-high/MYC-high/HLA-low tumors. Third, although Remodelin is a strong proof of concept, it is not yet a clinically established NAT10-targeted therapy in CRC. And finally, broader validation in additional in vivo and clinically relevant models would strengthen the path toward translation.

Overall, the translational message from our work is that targeting NAT10 may provide a practical way to break a MYC-driven, tumor-selective autophagy program, restore MHC-I expression, enhance T-cell infiltration, and sensitize immune-cold MSS colorectal cancer to checkpoint blockade.