Defining Therapeutic Epitranscriptome of Multiple Myeloma for Accurate Subtyping and Personalized Prognostics

Background: Multiple Myeloma (MM) is a heterogeneous disease, with risk classification (standard risk SR and high risk HR) mostly defined by cytogenetic changes. RNA covalent modifications (collectively addressed as the epitranscriptome) have recently emerged as key signatures of cancer progression, with RNA modification writer and eraser inhibitors being clinically trialed. For example, inhibitor of FTO, an eraser for N6-methyladenosine (m6A), was shown to synergistically suppress MM for mouse models in combined treatment. However, we still lack accurate, nucleotide-resolved landscape of the main RNA modifications, such as m6A, 5-methylcytosine (m5C), N4-acetylcytosine (ac4C) and pseudouridine (pU) in MM, which substantially restricts refined diagnostic, prognostic and drug development opportunities. Here we provide this missing information by investigating the epitranscriptomic signatures of SR and HR MM.

Methods: To obtain the isoform-resolved information about concurrent presence of m6A, m5C, ac4C and pU in native MM RNA at single-molecule, single-nucleotide resolution transcriptome-wide, we employed nanopore-based direct RNA sequencing (DRS). DRS was performed on purified tumor cells from three SR and three HR MM patients. Data were analyzed for the modification sites and stoichiometry using in-house developed SWARM software. Modifications were filtered based on probability thresholds controlling the false discovery rate. MM genetic markers and cytogenetic subtypes were analyzed for the epitranscriptome asscociations.

Results: For all modification types with the exception of m5C in standard-risk patients, we observed a significantly higher modification rate in the light and heavy chains of the immunoglobulin (Ig) genes compared to other mRNAs: m6A was increased by a 3.6 fold (σ 1.7) in HR patients and 1.4 fold (σ 0.7) in SR; m5C by a 5.4 (σ 6.4) and 0.71 (σ 0.1) fold; pU by a 4.2 (σ 3) and 3 (σ 1.4) fold; and ac4C by a 8.1 (σ 2.7) and 2.8 (σ 3) fold, respectively. While it is known that the variable segment of the Ig chains are highly prone to mutations, we show here that the constant regions instead have a significantly higher rate of modification, regardless of the chain type. More specifically, ac4C and m5C rates are the highest in Ig light chains, in particular in the constant region. In HR MM, m6A and pU are the highest in the Ig light constant segments while in SR MM patients, they are the highest in the Ig variable light segments.

Importantly, other MM prognostic genes such as B2M or JUND also globally exhibited a higher modification rate compared to the background mRNA average (m5C 5.6 and 9.6; ac4C 2 and 9.6; m6A 4.1 and 24; and pU 1 and 2.2 fold higher, respectively). Yet across all tested RNA, SR patients were characterized by a higher m5C and ac4C modification rate (compared to HR patients, ~4 fold) with nonetheless lower ac4C stoichiometry in modified sites, and a lower m6A and pU rate (~2 fold). HR patients displayed a higher expression of the main modification writer enzymes. Interrogation focused on ribosomal (r)RNA further indicates specific sites of differential modification stoichiometry that have previously been linked to reduced tumor-suppressor expression and aggressive tumor phenotypes in other cancer types.

Conclusion: Our data describe a diverse and rich epitranscriptomic landscape in Multiple Myeloma, strongly associating with the MM subtypes and risk. Association of poor prognostic with higher modification deposition together with specific rRNA modification pattern differences indicate that the epitranscriptome plays an important role in the biology of Multiple Myeloma, and suggest that epitranscriptome-resolved DRS data will provide more accurate subtyping and reveal new pathways and targets of drug response.