Deconstructing Regenerative Medicine: From Mechanistic Studies of Cell Therapy to Novel Bioinspired RNA Drugs

Abstract

All FDA-approved noncoding RNA (ncRNA) drugs (n=~20) target known disease-causing molecular pathways by mechanisms such as antisense. In a fortuitous evolution of work on regenerative medicine, my coworkers and I inverted the RNA drug discovery process: first we identified natural disease-modifying ncRNAs, then used them as templates for new synthetic RNA drugs. Mechanism was probed only after bioactivity had been demonstrated. The journey began with the development of cardiosphere-derived cells (CDCs) for cardiac regeneration. While testing CDCs in a first-in-human trial, we discovered they worked indirectly: ncRNAs within CDC-secreted extracellular vesicles (EVs) mediate the therapeutic benefits. The vast majority of such ncRNAs are fragments of unknown function. We chose several abundant ncRNA species from CDC-EVs, synthesized and screened each of them in vitro and in vivo. Those with exceptional disease-modifying bioactivity inspired new chemical entities that conform to the structural conventions of the FDA-approved ncRNA armamentarium. This discovery arc—Cell-Derived RNA from EVs for bioinspired Drug develOpment, or CREDO—has yielded various promising lead compounds, each of which works via a unique, and often novel, mechanism. The process relies on emergent insights to shape therapeutic development. The initial focus of our inquiry—CDCs—are now themselves in phase 3 testing for Duchenne muscular dystrophy (DMD) and its associated cardiomyopathy. But the intravenous delivery strategy and the repetitive dosing protocol for CDCs, which have proven key to clinical success, both arose from systematic mechanistic inquiry. Meanwhile, emergent insights have led to multiple cell-free therapeutic candidates: CDC-EVs are in preclinical development for ventricular arrhythmias, while the CREDO-conceived RNA drugs are in translation for diseases ranging from myocarditis to scleroderma.

Although this article is written in the first person, the work described is a group effort. The online supplement provides a comprehensive list of my coauthors over the last two decades. I gratefully share credit with these exceptionally talented individuals (see data supplement).

In 2004, while I was Editor of Circulation Research, I became interested in regenerative medicine. This seemingly random career inflection—I am a cellular electrophysiologist by training—arose from prior work in my lab using human embryonic stem cells (hESCs) as biological pacemakers¹. Working with hESCs got me thinking about cellular cardiomyoplasty and its game-changing potential for heart failure and other cardiac disorders. But practical application of hESCs for regenerative applications seemed elusive: such cells are immunogenic, they form tumors, and the cardiomyocytes they produce are immature. Thus, I pivoted to non-pluripotent cells as therapeutic candidates. My coworkers and I produced and characterized an intriguing stromal/progenitor cell type from the adult heart—cardiosphere-derived cells, or CDCs—which we found to have striking cardioprotective and immunomodulatory benefits in several disease models, and disease-modifying bioactivity in clinical trials. CDCs are now in phase 3; no other cell type tested for heart disease remains in realistic contention for FDA approval. Key to success has been following the data, especially when the data negated our preconceptions. The most surprising part of this journey has been the unexpected destination—novel, synthetic RNA drugs, bioinspired by the contents of extracellular vesicles (EVs) secreted by CDCs. It is that journey of discovery that I will review here.

CDCs.

Since my lab first reported the isolation of CDCs in 2007², >300 papers have appeared on this cell type, from >80 independent labs worldwide. The name CDC reflects the manufacturing process, which is depicted in Fig. 1A. In primary culture, human cardiac tissue explants shed cells that self-organize into spherical clusters called cardiospheres. We then replate cardiospheres on adherent cultureware to yield CDCs. CDCs qualify as cardiac progenitor cells: they are of intrinsic cardiac origin³, multipotent and clonogenic⁴. CDCs uniformly express CD105 and are negative for CD45 and other hematogenous markers. A number of clinical trials, listed in Fig. 1B, have used, or are using CDCs⁵. Three other trials of CDCs not listed here have been conducted independently in Japan, in hypoplastic left heart syndrome and in heart failure^6–8. (For a review of other clinical trials using diverse cell types to treat heart disease, see ref⁹.). The initial rationale was that CDCs would work canonically, i.e., engraft, proliferate, and differentiate into new myocardium (as reviewed previously¹⁰); consistent with this presumption, we developed autologous CDCs as the therapeutic candidate for the first-in-human CADUCEUS trial¹¹. While CADUCEUS was under way, however, we discovered that, despite the fact that CDCs meet the criteria for being progenitor cells, they don’t work that way. Few (<<1%) of the injected CDCs—whether syngeneic or allogeneic—are measurable 3–4 weeks after transplantation, but functional and structural benefits persist for at least 6 months^12,13. During the ~2 weeks that appreciable numbers of transplanted CDCs persist in the tissue, CDCs indirectly induce cardiomyogenesis and immunomodulation in the host myocardium, promoting structural and functional recovery¹⁴. These observations prompted us to shift from autologous to allogeneic, banked CDCs for all the clinical trials after CADUCEUS. Attempts to pinpoint the underlying paracrine signals released by CDCs initially focused on growth factors such as SDF-1¹². The emphasis shifted in 2014, when we first implicated exosomes as the mediators of CDC benefits¹⁵.

Figure 1. CDCs.

A: Process flow for CDC manufacturing from human heart tissue. Chunks of cardiac tissue, typically 10–50 mg in mass, are minced into “explants” each about the size of a grain of sand. The cells that spontaneously grow out from such explants in culture (explant-derived cells) are harvested and replated on non-adherent cultureware, where they self-assemble into cardiospheres. Such cardiospheres are themselves replated onto adherent culture plasticware to grow CDCs. B: CDC clinical trials in the USA. Trial names, design and indication on the left; delivery and dosing summary on the right. Abbreviations: RPCT: randomized prospective clinical trial. MI: myocardial infarction. HFrEF: heart failure with reduced ejection fraction. RDBPCT: randomized double blind placebo-controlled trial. HFpEF: heart failure with preserved ejection fraction. IV: intravenous.

EVs.

First, a comment on nomenclature: the term EVs includes all extracellular lipid bilayer vesicles; exosomes are a particular type of EV, ~100 nm in diameter, which are products of the Golgi fused with multivesicular bodies and rich in RNA contents (see Fig. 2A for summary of exosome biology). Although the therapeutic bioactivity of CDCs is primarily attributable to exosomes, we prefer the generic term CDC-EVs, as it does not promise absolute homogeneity of the EV population under study. Serum-free media conditioned by CDCs contains EVs which include many exosomes, as gauged by particle size and antigenic profile. In a mouse model of acute myocardial infarction (MI), blocking exosome production renders CDCs ineffective (Fig. 2B). Meanwhile, CDC exosomes on their own reproduce the benefits of the parent CDCs (Fig. 2C)¹⁵. Thus, exosomes are necessary and sufficient for CDC efficacy. The concepts articulated here are not limited to CDCs; even pluripotent stem cell products are now acknowledged to act, at least partially, via exosome secretion^16–18. As suggested by the data in Fig. 2C, CDC-EVs may themselves be attractive next-generation cell-free therapeutic candidates. Indeed, we are currently performing IND-enabling studies with a view to first-in-human clinical application of CDC-EVs in patients with recurrent ventricular tachycardia¹⁹. We and others have further established that many, if not most, of the effects of exosomes are mediated by their RNA contents, specifically noncoding RNAs (ncRNAs²⁰). The emerging paradigm is as follows: CDCs secrete exosomes which transfer ncRNA payloads into target cells, inducing transcriptomic and phenotypic changes that underlie the benefits of CDC therapy. It is noteworthy that CDC-EVs, while derived from the heart, affect many other tissues: as but one example, CDC-EVs increase skeletal muscle regeneration and force production, mimicking the effects of systemic CDC infusion in a mouse model of DMD²¹. These findings motivated the repeated intravenous dosing paradigm²² now in phase 3 testing, with the hope of delaying loss of upper limb function (and progression of cardiomyopathy) in boys and young men with advanced DMD.

Figure 2. Exosomes, a specific population of EVs.

A. Schematic summary. Exosome biogenesis features fusion of surface membrane invaginations with products of the Golgi. When such multivesicular bodies fuse with the surface membrane, they discharge EVs into the extracellular space. LncRNA: Long noncoding RNA. Inset provides bullet highlights of exosome biology. B. Blockade of CDC exosome biosynthesis with GW4869 (a small molecule inhibitor of neutral sphinogomyelinase) abrogates, while (C.) exosomes isolated from CDC-conditioned media mimic, the functional benefits of CDCs in a post-MI mouse model. LVEF: left ventricular ejection fraction; CTRL: vehicle injection only; CDC-XO: CDC exosomes; NHDF-XO: exosomes from normal human dermal fibroblasts. B,C from Ibrahim et al, 2014¹⁵.

RNA contents.

EV populations derived from different cell types have quite distinctive RNA payloads²³. Unbiased RNA sequencing of CDC-EV cargo reveals >10,000 molecularly distinct ncRNA entities, most of which have no known bioactivity²⁴. Fig. 3A&B show the diversity of RNA types in CDC-EVs. (We focus on RNA as EV bioactivity is virtually eliminated when the contents are exposed to RNase, and on ncRNA, as intact messenger RNAs are relatively rare in CDC-EVs²⁴.) The diverse cargo creates opportunities for identifying new defined factors as therapeutic candidates. Focusing on species 20–200 nucleotides in length, we identified exceptionally plentiful ncRNAs of unknown function (Fig. 3C) and screened them individually for bioactivity. Selected species were synthesized by solid phase chemistry, coated in DharmaFECT^® transfection reagent, and exposed in vitro to macrophages and/or lymphocytes in primary culture. Using recipient cell transcriptomics as the readout, species that induced salutary changes in gene expression (e.g., activation of anti-inflammatory and/or anti-fibrotic gene pathways) were further characterized in vivo in preclinical disease models, most often a rodent model of MI. Y RNAs are of particular interest as they are abundant in CDC-EVs (18% of small RNAs, Fig. 3A), and poorly characterized. EVs from various sources contain products of the four human Y RNA genes, but these ncRNAs are rarely full-length, and their function (if any) was unknown²⁵.

Figure 3. CDC-EVs and their cargo.

A,B. RNA sequencing of CDC-EV cargo shows Y RNA to be plentiful (second only to tRNA). C. Histogram of the abundance of the 21 most abundant RNA fragments in CDC-EVs. Adapted from L. Cambier et al., 2017²⁴.

The single most plentiful ncRNA species in CDC-EVs, EV-YF1, is a 56-nucleotide (nt) molecule encoded by the human YRNA4 gene (Fig. 4A). In vitro, EV-YF1 increases macrophage IL10 secretion (Fig. 4B); when infused in vivo, EV-YF1 is cardioprotective against MI (Fig. 4C,D)²⁴. We went on to report striking benefits in a model of hypertension and hypertrophy induced by angiotensin II infusion²⁶, where EV-YF1 exerted antifibrotic and anti-hypertrophic effects comparable to those of CDC-EVs. Intravenous EV-YF1 is also highly effective in a transgenic mouse model of hypertrophic cardiomyopathy²⁷. Thus, an EV-derived short Y RNA of unknown function turns out to have broad disease-modifying bioactivity.

Figure 4. EV-YF1 sequence and therapeutic effects.

A: Alignment of human Y RNA4 (hY4) gene with EV-YF1. The T insertion at position 15 turned out to be an artifact; the actual sequence is a perfect homologue of the first 56 nucleotides of hY4. Lower panel: When synthesized and packaged in a transfection reagent, EV-YF1 stimulates IL-10 expression in rat macrophages (B) and reduces infarct mass in rats (C,D). Ys: a control oligonucleotide with the same nucleotide content as EV-YF1 but scrambled in sequence. Adapted from L. Cambier et al., 2017²⁴.

From native RNAs to bioinspired new chemical entities.

The case study of EV-YF1 illustrates the power of mining EVs for bioactive ncRNAs. But natural ncRNAs are not ideal drug candidates: they are susceptible to degradation by RNase, and they do not conform to the structural conventions established by the clinical ncRNA armamentarium^28–30. As a case in point, EV-YF1 is 56 nt and unmodified, while all FDA-approved ncRNA drugs (of which there are now ~20) are 20–35 nt in length and chemically-modified for increased stability and decreased immunogenicity. To overcome this limitation, we chose the most promising natural ncRNAs and used them as bioinspiration for new chemical entities (NCE) that conform to ncRNA drug structural norms. Each native structure served as a template for structure-activity relationship optimization; resultant NCE were further tested in vitro and in vivo.

Discovery-based drug development.

Our drug development process inverts the usual ncRNA target-based paradigm. As illustrated in Fig. 5, the conventional method (upper panel) first identifies a disease pathway, based on a known mechanism, then targets it with an ncRNA that will inhibit the pathogenic process³¹. Patisiran, one such drug, is a small interfering RNA designed to inhibit the synthesis of transthyretrin (ttr), whose accumulation has been implicated in ~30% of amyloidosis cases. Less production of ttr halts excessive accumulation of the pathogenic protein in the heart and other tissues. In contrast, our discovery-based paradigm (Fig. 4, lower panel) begins by mining therapeutic EVs for abundant but obscure ncRNAs. If the natural ncRNA turns out to be bioactive, it then serves as bioinspiration for NCEs. Those with exceptional disease-modifying bioactivity undergo further development (dose optimization, toxicology studies etc.) as therapeutic candidates. The mechanism of benefit starts as a complete mystery, which we tackle only after we are convinced of the novel molecule’s therapeutic potential. This approach—CREDO—has yielded various promising lead compounds, each of which works via a unique, and often novel, mechanism.

Figure 5. Conventional ncRNA drug development method (upper panel) versus discovery-based paradigm (CREDO).

See text for details.

New lead compounds with disease-modifying bioactivity.

Using the CREDO paradigm, we have discovered and characterized a number of NCE lead compounds (chemically-modified mutant ncRNAs, each 24–35 nucleotides in length), which are in various stages of translation:

1. TY1, derived from EV-YF1. TY1 targets macrophages and acts in a novel manner by attenuating the DNA damage response³². Disease-modifying bioactivity has been demonstrated in animal models of MI, scleroderma with systemic sclerosis³³, and heart failure with preserved ejection fraction³⁴. Profound anti-fibrotic effects, with improvements in functional capacity, are evident in diverse preclinical models studied to date, including rodents and pigs.

2. TY2, derived from yREX3, a 3´ fragment of the human YRNA4 gene (non-overlapping with EV-YF1). TY2 acts by suppressing Pick1 and enhancing macrophage efferocytosis—an unprecedented mechanism. Disease-modifying bioactivity has been demonstrated in animal models of MI, where the cardioprotective benefits are manifested as major reductions of infarct size and circulating troponin I levels³⁵.

3. TT1, derived from tREX1, a bioactive fragment of a human tRNA that codon-matches glutamate. The mechanism of TT1 is as-yet unclear. Disease-modifying bioactivity has been demonstrated in the mdx mouse model of DMD³⁶, where disease progression is attenuated or even reversed, and in a rat model of MI.

4. TL1, TL2, TL3, derived from long noncoding RNA BCYRN1. The TL compounds are synthetic microRNA sponges that enhance the proliferation and activation of regulatory T cells. Disease-modifying bioactivity has been demonstrated in a rat model of MI, in which the cardioprotective effects correlate well with enhanced levels of circulating regulatory T cells³⁷.

Fig. 6 summarizes the various lead compounds and their current state of development. TY1, a derivative of EV-YF1 (referred to in Fig. 6 as yREX-1; see ref²⁶ for a primer on nomenclature), is the furthest along in translation, having already entered the FDA approval pathway at the preIND level.

Figure 6. New ncRNA drugs bioinspired by CDC-EV cargo.

From left to right, columns show: ncRNA family of the native RNA; names of the natural entities; target cell type and cellular changes with drug exposure; disease-modifying bioactivity (DMA); mechanism of action (MoA); names of synthetic NCEs and disease models of verified bioactivity. Abbreviations: MI, myocardial infarction; HCM, hypertrophic cardiomyopathy; DCM, dilated cardiomyopathy; DMD, Duchenne muscular dystrophy; TREX1, 3´ DNA exonuclease; Pick1, Protein interacting with C kinase 1; SSc, systemic sclerosis; HFpEF, heart failure with preserved ejection fraction; miR, microRNA; Macs: macrophages; Tregs, regulatory T cells

Unexpected mechanistic insights.

The CREDO process is agnostic to mechanism; we only require that the successful candidate be novel, bioactive, and reducible to FDA-friendly structural conventions. Once bioactivity is established, however, the hunt is on for mechanism. This can be a laborious process, as typically nothing is known about the natural parent ncRNA other than its annotation to one ncRNA family or another, which can itself be unreliable³⁸. When mechanism is ascertained, it can be revelatory. As presented above and summarized in Fig. 5, 5 of the 6 NCEs studied so far work by previously-unrecognized mechanisms; the mechanism of the sixth NCE, TT1, is still unclear.

Summary.

The CREDO drug discovery paradigm—mining therapeutic EVs for bioactive ncRNAs as templates for regulatory-compliant NCEs—is potent and generalizable. How did this paradigm come about? As we proceeded in developing CDCs for regenerative medicine applications, our focus shifted from CDCs to EVs as potential treatments for heart failure post-MI. We then further shifted the focus to ncRNAs in EVs. The realization that EV cargo is richly diverse, and that specific factors are therapeutically active, further moved our attention from EVs to defined molecular entities. Inventories of therapeutic EV contents, generated by RNA sequencing, now serve as treasure chests to begin to decipher and exploit the complex biology of EVs. Defined ncRNAs mined from EVs can either be used as they occur in nature or serve as templates for the creation of NCEs with enhanced stability and potency. Bioinspired NCE such as TY1 have the potential to become synthetic, scalable, off-the-shelf drugs, complementary to the existing armamentarium. Once the mechanism is deciphered, we might discover that therapeutic potential is no longer limited to heart disease. TY1, which acts to attenuate DNA damage²⁶, serves as an example. Despite the cardiac origin of its parent ncRNA (EV-YF1), TY1 has anti-fibrotic activity not only in heart but also in a variety of non-cardiac tissues (including lung and skin in models of scleroderma with systemic sclerosis²⁷).

The journey has been a circuitous one. Why did I choose “deconstructing regenerative medicine” for the title? “Deconstruction” is full of nuance for linguists and philosophers, who, for centuries, have labored under that rubric to debate the very meaning of language³⁹. Here, I simple-mindedly followed the Merriam-Webster definition of deconstruction: “the analytic examination of something (such as a theory) often in order to reveal its inadequacy⁴⁰”. The process of discovery reviewed here has, indeed, revealed the inadequacy of the regenerative medicine dogma. We started by drinking the Kool Aid that stem/progenitor cells would implant in the injured myocardium, proliferate and differentiate, creating new cardiac tissue. While wholesale replacement of injured heart tissue by progenitor cells may be possible, it has yet to materialize, despite decades of effort. The results of my own mechanistic experiments on CDCs pointed in a different direction. If it works at all, cardiac cell therapy (at least as applied in its first three decades) does not work according to canonical stem cell teachings. The benefits are indirect and mediated by EV secretion. Nevertheless, the outcome of this journey—ncRNA drugs with untold therapeutic potential—is a fortuitous consequence of cell therapy, dissected mechanistically. We have ended up with multiple defined molecular entities of broad therapeutic potential—a testament to the serendipitous power of open-ended scientific inquiry.