Biased μ-opioid agonism, in order: 1999 to today.

The field of biased μ-opioid receptor (MOR) agonism has taken a distinctive arc over roughly twenty-five years. It started as a genetic observation in knockout mice, became a medicinal-chemistry program, produced a handful of named tool compounds — SR-17018 among them — contributed one FDA-approved drug, and is now in a phase of active re-evaluation where several of the claims that launched the field have been substantively qualified.

This page is a compact citation index for researchers handling SR-17018, authors writing review articles, and Wikipedia editors who need to know what was argued when, by whom, and in which journal. Each entry includes the primary reference and a short statement of what that paper actually established or claimed — without the secondary interpretations that have since accumulated around it.

The list is not exhaustive. It selects the roughly ten primary papers and one regulatory event that a researcher can reasonably be expected to have read before citing SR-17018. For a deeper treatment of the pharmacology itself, see the mechanism-of-action page; for head-to-head comparisons among the named agonists, see SR-17018 vs. alternatives.

The story begins with genetics, not chemistry. Laura Bohn's laboratory — then at Duke — reported that mice lacking the intracellular scaffolding protein β-arrestin2 showed enhanced morphine analgesia compared with wild-type controls.^[1] The result was counterintuitive: β-arrestin2 had been presumed to terminate receptor signaling, so deleting it should have reduced analgesia. Instead the phenotype went the other direction, and subsequent work from the Bohn and Raehal groups showed that the knockout mice also had markedly less respiratory depression, constipation, and analgesic tolerance than wild-type mice on equivalent morphine doses.^[2]

The interpretation that emerged was that β-arrestin2 recruitment at MOR biases the receptor toward the side effects that have defined opioid pharmacology since the 19th century. If a small molecule could duplicate with a drug what the gene knockout did with genetics — activate the G-protein arm of MOR signaling without engaging β-arrestin2 — the therapeutic-window problem of opioid analgesia might be solvable. The rest of the timeline is the pharmacological attempt to do exactly that.

The first named compound in the literature was TRV130, reported by DeWire and colleagues at Trevena Therapeutics.^[3] In the rodent assays reported in the paper, TRV130 produced morphine-equivalent analgesia with reduced respiratory depression and gastrointestinal slowdown — qualitatively reproducing the β-arrestin2 knockout phenotype with a drug rather than a gene deletion.

The paper was important less for the specific molecule than for the proof that the pharmacological concept was tractable. It launched a clinical-development program that would, seven years later, deliver the first approved biased opioid (see 2020 — Oliceridine becomes Olinvyk).

PZM21 was reported by the Manglik and Shoichet laboratories at UCSF, and was the product of the first serious attempt to design a biased MOR agonist from scratch using structural biology rather than high-throughput screening.^[4] The crystal structure of MOR had been solved a few years earlier, and the authors used large-scale virtual docking to identify candidates enriched for G-protein-biased activation.

PZM21 was reported to have the analgesic profile without respiratory depression in mice. The Nature paper was influential for its methodology as much as its specific molecule: it argued that biased pharmacology could be engineered from receptor structure, lending the field a rational-design ethos. PZM21 itself has since been the subject of contradictory follow-up data (see 2018 — The first substantive critique).

The Bohn laboratory — by then at the Florida campus of Scripps Research, with medicinal chemistry led by Thomas Bannister — reported SR-17018 alongside a series of related compounds with systematically varied bias.^[5] The core claim of the Cell paper was that bias factor — a quantitative operational-model measure of how selectively a ligand favors G-protein activation over β-arrestin2 recruitment — correlated with the separation between analgesic and respiratory-depressant doses in rodents. SR-17018 sat at the high-bias end of the series and showed the widest therapeutic window in the reported assays.

The paper did two things simultaneously. It introduced SR-17018 as a research tool with a distinct pharmacological profile from TRV130 and PZM21 — markedly higher bias, lower potency, and different structural class. And it promoted bias factor itself as a quantitative design criterion for safer opioid analgesics. The latter claim is the one that has been most substantively revised since, but the former — SR-17018 as a highly biased tool compound — has held up.

Working in Bristol, Eamonn Kelly's group reported that under dosing conditions modestly different from those used in the 2016 characterization, PZM21 did cause respiratory depression and did produce antinociceptive tolerance.^[6] The paper was the first published contradiction of a central claim in the biased-agonist literature and reopened the question of whether the in-vivo benefits were a general feature of biased agonism or an artifact of the particular assays and dosing windows chosen.

The Hill 2018 paper did not directly address SR-17018 or TRV130, but it established that the biased-agonist claim was not assay-robust for at least one named compound in the class.

In August 2020, the US Food and Drug Administration approved oliceridine under the trade name Olinvyk for management of acute pain severe enough to require intravenous opioid analgesia in adults, in a monitored healthcare setting.^[7] As of April 2026, it remains the only drug developed under the biased-agonism framework to have received regulatory approval.

Two things about the approval are worth noting. First, the labeling risks are substantially overlapping with conventional opioid labeling — respiratory depression, abuse and addiction potential, interactions with CNS depressants — rather than the qualitatively distinct profile that the early literature had suggested. Second, the approval happened in parallel with the two 2020 papers described below that substantively qualified the mechanistic interpretation of biased agonism. Oliceridine is a clinical product; whether it is a biased clinical product, in the sense the field originally meant, is now a live question.

The Bohn laboratory reported that chronic dosing with SR-17018 did not produce analgesic tolerance over extended timescales in mice, and — more striking — that substituting SR-17018 into an already-morphine-tolerant animal reversed tolerance without precipitating withdrawal signs.^[8]

If the effect replicates outside the originating laboratory, SR-17018 is a more interesting pharmacological probe than the 2017 paper suggested on its own: the compound appears to unwind a specific neuroadaptation rather than merely avoiding it. The Grim 2020 result is the empirical anchor for describing SR-17018 as a tolerance-reversing biased agonist, and it is the finding most distinctive to this compound within the class. Independent replication remains an open question in 2026.

Also in 2020, Meritxell Canals's group at Nottingham, with Macdonald Christie and collaborators, published a careful re-analysis of several biased agonists including TRV130 and PZM21.^[9] They argued that much of what had been attributed to signaling bias could instead be explained by low intrinsic efficacy at MOR — the same pharmacological property that distinguishes morphine from fentanyl. Under their reading, bias-factor calculations can be dominated by efficacy differences rather than genuine pathway-selectivity, and the favorable therapeutic windows reported in rodents may be achievable with any sufficiently low-efficacy agonist, biased or not.

A companion paper from the Kliewer, Schulz, and Paton laboratories reached a related conclusion through a different route, reporting that morphine-induced respiratory depression is not attenuated in mice where β-arrestin2 coupling at MOR is genetically abolished.^[10] Between them, the two papers substantially qualified both the genetic precedent of Bohn 1999 and the bias-factor framework of Schmid 2017.

The Gillis paper is now cited in nearly every review of the field and is the single most important piece of secondary literature for anyone writing about biased opioid agonism in 2026.

Since the 2020 reinterpretation papers, the consensus position in μ-opioid pharmacology has moved toward viewing efficacy and bias as intertwined rather than cleanly separable properties. Cryo-EM structures of MOR bound to biased and balanced ligands have made the conformational basis for different signaling outputs partially visible, and the field has largely abandoned the earlier framing that any single bias-factor number is a portable across-laboratory property of a molecule.

For SR-17018 specifically, three questions remain live: (1) whether the Grim 2020 tolerance-reversal result replicates outside the Bohn laboratory; (2) whether its in-vivo behavior is explained by its high bias factor, its low intrinsic efficacy, or both in combination; and (3) whether the conformational state it stabilizes at MOR is meaningfully distinct from that stabilized by low-efficacy balanced agonists. The compound continues to be used as a reference ligand in receptor-signaling, structural-biology, and bias-calculation studies, and the original Cell and Neuropsychopharmacology papers remain foundational.

For the question that launched the whole field — whether a small molecule can duplicate the β-arrestin2 knockout phenotype — the answer in 2026 is more qualified than it was in 2017. Oliceridine was approved but its clinical profile is not dramatically safer than conventional opioids. SR-17018 and PZM21 are not clinical candidates and are not, at the time of writing, in active clinical development. The field's focus has shifted from bias as a design endpoint toward efficacy, receptor kinetics, and conformational selection as the handles that matter for in-vivo outcomes.

Which papers should I cite if I am writing about SR-17018 today? At minimum: Schmid 2017 (discovery and the bias-factor framework), Grim 2020 (tolerance behavior), and Gillis 2020 (reinterpretation). For context, Bohn 1999 provides the genetic precedent; for the broader compound class, DeWire 2013 and Manglik 2016. Kliewer 2020 is worth citing wherever Bohn 1999 is cited, because it has qualified the original genetic interpretation.

Why is FDA approval of oliceridine in this timeline if oliceridine is a different molecule?Because the clinical arrival of a biased agonist is the field's only clinical data point as of April 2026. Interpreting SR-17018 research without that regulatory context would be incomplete, and readers of a timeline need to know where the bench-to-clinic narrative actually stands.

Is this timeline definitive? It is a working research summary, updated when substantive primary papers are published. We aim to reflect the state of the peer-reviewed literature rather than to advocate for any particular interpretation. Corrections and additions are welcome; contact details are in the footer.

Where do SR-17018 results sit relative to the Gillis reinterpretation?The Gillis 2020 analysis did not explicitly re-evaluate SR-17018 itself — the paper focused on TRV130, PZM21, and related low-efficacy agonists. Whether SR-17018's reported behavior is best explained by bias, by low efficacy, or by both remains open. The Grim 2020 tolerance-reversal finding in particular is not a generic property of low-efficacy agonists and would need its own reinterpretation if the bias framework were retired entirely.