SR-17018 vs. TRV130, PZM21, and morphine — what the literature actually shows.

More recent data narrowed the gap between the early “bias predicts safety” narrative and the alternative view that low intrinsic efficacy explains much of the observed profile.

Every claim that has ever been made about SR-17018 — that it is highly biased toward G-protein signaling, that it has a wider therapeutic window than morphine, that it reverses morphine tolerance — is a comparative claim. It only means anything relative to a reference compound, run in the same assay, in the same laboratory, on the same day. Reading the biased-agonism literature without keeping that in mind is how researchers end up disappointed by their own data.

This page walks through the comparison that the Bohn lab drew in the 2017 Cell paper that introduced SR-17018^[1] and the comparisons that other groups have drawn since, across four compounds that anchor most of the preclinical work being published today: morphine (the balanced full agonist everyone calibrates against), TRV130(oliceridine, Trevena's biased agonist, the only member of this family to reach the clinic),^[4] PZM21 (Manglik lab, a structure-based design from an opioid-receptor docking campaign),^[5] and SR-17018 itself.^[1]

Each pairing answers a slightly different question. SR-17018 vs. morphine asks whether a highly biased tool compound actually separates analgesia from respiratory depression in rodents. SR-17018 vs. TRV130 asks whether a more biased ligand does meaningfully better than a mildly biased one that survived clinical development. SR-17018 vs. PZM21 asks whether two differently-biased compounds behave differently in-vivo. And the full set, seen together, asks whether the signaling-bias framework itself is predictive — a question to which the answer is now less clean than it seemed in 2017.

Before drilling into specific comparisons, a one-paragraph sketch of each compound and why it matters.

Morphine. The prototypical μ-opioid agonist, isolated in 1804, still the reference against which every new opioid is characterized. Morphine recruits both G-protein and β-arrestin2 signaling at MOR — it is the archetypal balanced agonist in these datasets. It produces analgesia, respiratory depression, constipation, and tolerance on chronic dosing. In the rodent assays used for biased-agonism work, morphine sits at the narrow-therapeutic-window end of every comparison.

TRV130 (oliceridine). Synthesized by Trevena and first reported by DeWire and colleagues in a 2013 paper in JPET.^[4] TRV130 was designed as a G-protein-biased MOR agonist and reported to produce analgesia with reduced respiratory depression and gastrointestinal slowdown relative to morphine in preclinical models. It advanced through Phase III clinical trials and was approved by the FDA in 2020 under the trade name Olinvyk for use in controlled hospital settings. Its bias factor, as recomputed in the Schmid 2017 dataset, is modest — lower than SR-17018, higher than morphine.^[1]

PZM21. Discovered through a structure-based virtual screen against a high-resolution crystal structure of MOR, and reported by Manglik and colleagues in a 2016 Nature paper.^[5] The original characterization described PZM21 as an analgesic with greatly reduced respiratory depression and no morphine-like reward in rodents. A 2018 follow-up from the Henderson lab in British Journal of Pharmacology reported that PZM21 does cause respiratory depression and does produce antinociceptive tolerance under conditions that differ only modestly from the original paper, substantially tempering the initial claims.^[6]

SR-17018. Reported by Schmid and colleagues in the 2017 Cell paper that introduced the quantitative bias-factor framework.^[1] Sits at the high-bias end of that dataset, with a reported therapeutic window substantially wider than morphine's. A 2020 follow-up from the same group reported that SR-17018 reverses morphine tolerance in mice without precipitating withdrawal.^[2] Has not entered clinical development.

Bias factor is a quantitative measure of how selectively a ligand activates one intracellular pathway over another at the same receptor. In the MOR literature, the two pathways compared are virtually always G-protein coupling (measured by Gαᵢ activation, cAMP inhibition, or GTPγS binding) and β-arrestin2 recruitment (measured by bioluminescence resonance energy transfer, or BRET, or by PathHunter-style complementation assays).

The calculation rests on the operational model of pharmacological agonism. For each pathway, a ligand's concentration-response curve is fit to extract a transduction coefficient (log τ/Kₐ), which captures how efficiently that ligand produces a response given its affinity for the receptor. The difference between the two transduction coefficients, normalized to a reference ligand, is the bias factor. By convention, DAMGO is set to zero bias, and the sign indicates direction.

In the Schmid 2017 dataset, run with a consistent set of assays in one laboratory, the ordering that emerged was roughly: morphine ≈ DAMGO (near zero bias) < TRV130 (modest G-protein bias) < SR-17018 (strong G-protein bias), with SR-17018 standing out as the most cleanly biased compound in the series.^[1] PZM21, although not in the Schmid series, has been characterized in independent datasets as having modest G-protein bias comparable to TRV130.^[5]

Two cautions are essential here. First, bias factors are not portable across laboratories. The absolute values shift depending on cell line, receptor expression, coupling efficiency, and which assay is chosen for each arm. A compound that looks highly biased in one lab can look modestly biased in another. When comparing SR-17018 with TRV130 or PZM21, use datasets where all three compounds were run side by side, not numbers pulled from different papers.

Second, the sign and magnitude of bias depend on the reference ligand chosen. DAMGO is conventional; using morphine as reference, which some groups prefer because morphine is closer to the clinically relevant space, shifts all the numbers. Read methods sections carefully before interpreting any headline number.

The most important in-vivo comparison in this literature is the gap between the dose that produces analgesia and the dose that produces respiratory depression. A compound with a wider gap is, in principle, safer — overdose is less likely at any given analgesic dose. The Schmid 2017 paper made this the central claim of the biased-agonism framework: that bias factor, measured in cell-based assays, predicts this gap in mice.^[1]

In the rodent datasets reported in the 2017 paper, the ordering was consistent with that prediction. Morphine showed the narrowest window — respiratory depression began at doses only modestly above those producing analgesia. TRV130 was intermediate. SR-17018 showed the widest separation of the series, with hot-plate and tail-flick analgesia at doses where plethysmographic measures of minute ventilation were essentially unaffected.

PZM21, in its original characterization, also showed a notably widened window relative to morphine.^[5] The Hill 2018 paper, however, reported that this result was sensitive to methodology: when respiratory measurements were made using plethysmography protocols closer to those used by other groups, PZM21 did cause meaningful respiratory depression.^[6] The methodological divergence has not been fully reconciled, but it is a cautionary example of how sensitive the therapeutic-window endpoint is to exact experimental conditions.

Reading these comparisons today requires holding two things at once. The ordering within any single-laboratory comparison is largely reproducible: SR-17018 has a wider window than morphine in every controlled comparison we are aware of. But the absolute window size, and the ordering between compounds characterized in different laboratories, is not stable.

Tolerance — the progressive loss of analgesic efficacy on repeated dosing — is the property that, if cleanly separated from analgesia, would transform the clinical utility of opioid therapy. The original β-arrestin2 knockout work from the Bohn laboratory established the animal-genetic basis for thinking tolerance and analgesia might be separable in principle: mice lacking β-arrestin2 retained morphine analgesia but developed less tolerance on chronic dosing.

Morphine, as the reference, produces robust tolerance in every standard rodent assay on chronic dosing schedules of a few days to two weeks. TRV130, in preclinical work, was reported to produce less tolerance than morphine, though the difference is not as striking as the respiratory-depression difference.^[4] PZM21 was initially characterized as producing little tolerance, but the 2018 Hill paper reported meaningful antinociceptive tolerance under standard chronic-dosing protocols, again tempering the original claim.^[6]

SR-17018 is the most striking compound in this dimension. The Grim 2020 paper reported that mice dosed chronically with SR-17018 retained analgesic efficacy on repeated dosing, with little or no development of tolerance over the tested window. More strikingly, when animals that had developed morphine tolerance were switched to SR-17018, not only did SR-17018 produce full analgesia, but subsequent morphine doses regained efficacy — i.e., SR-17018 appeared to reverse established morphine tolerance. Critically, this substitution did not precipitate the withdrawal signs that typically accompany switching to a less-efficacious opioid in a dependent animal.^[2]

If the Grim 2020 findings replicate robustly in other laboratories, SR-17018 is a more interesting pharmacological probe than the original 2017 paper suggested. As of the last review date at the top of this page, we are not aware of independent replication at the same scale, although multiple groups have published adjacent chronic-dosing work.

Of the four compounds, only two are available as clinical drugs. Morphine is a controlled substance in every jurisdiction we are aware of and is dispensed under prescription. TRV130, branded Olinvyk, was approved by the US FDA in 2020 for the management of acute pain severe enough to require an intravenous opioid analgesic in adults, restricted to controlled settings such as hospitals. TRV130 is a Schedule II controlled substance in the US. Its clinical reception has been modest; commercially it has not displaced the conventional opioid armamentarium.

PZM21 has not entered clinical development as far as we can determine from the published literature, and given the 2018 respiratory-depression follow-up, clinical advancement would require fresh preclinical support.^[6]

SR-17018 has not entered any human clinical trial we are aware of. It is supplied as a research-grade compound for in-vitro pharmacology and preclinical rodent studies. As discussed in the safety and handling page, it is not scheduled under the US Controlled Substances Act or the major international research- chemical schedules we track as of the review date. Regulatory status varies by jurisdiction and changes faster than our page reviews, so confirm before receipt.

The table below summarizes status. It is not exhaustive — local regulations may add restrictions that are not visible at the federal or international level.

Any comparison drawn in 2026 that does not acknowledge the Gillis 2020 Science Signaling paper is incomplete. The paper, from the Canals and Christie groups, argued that the in-vivo advantages reported for biased agonists can be explained parsimoniously by low intrinsic efficacy at MOR, not by signaling bias per se.^[3]

The logic is straightforward. In cell-based assays, β-arrestin2 recruitment typically requires higher receptor occupancy than G-protein coupling — the arrestin arm is effectively “downstream” in the sense that it takes stronger ligand-receptor interaction to drive it. A low-efficacy ligand, which cannot fully activate the receptor, will therefore look biased toward the easier-to-drive pathway (G-protein activation) simply because it can partially drive that pathway but not the harder one (β-arrestin2 recruitment). Under this reading, “bias” and “low efficacy” are empirically entangled, and much of the in-vivo therapeutic window observed with biased agonists could be attributable to the same property that distinguishes morphine (lower efficacy) from fentanyl (higher efficacy).

The Kliewer 2019 Nature Communications paper, using phosphorylation-deficient MOR knock-in mice that cannot recruit β-arrestin2 at the receptor, found that loss of β-arrestin2 recruitment improved analgesia and reduced tolerance but actually worsened respiratory depression and constipation — the opposite of what the bias hypothesis predicts.^[7]That result is hard to reconcile with a simple “β-arrestin2 causes the bad side effects” model and has pushed the field toward the intrinsic-efficacy framing.

How does this affect the four-way comparison on this page? If Gillis 2020 is correct, the comparisons presented in earlier sections are still valid in their empirical details — hot-plate potencies, plethysmographic responses, tolerance profiles — but the mechanistic interpretationshifts. SR-17018's wider therapeutic window may reflect low intrinsic efficacy as much as, or more than, signaling bias. That does not make the compound less useful as a tool or as a potential template for safer opioid design; it means the right story about why it works is less settled than the 2017 paper suggested.

If you are designing a rodent assay, a BRET experiment, or a receptor-kinetic study that includes SR-17018, the question of which other compounds to run is a question about what claim you want to support or refute. Three common cases:

• You want to benchmark SR-17018's therapeutic window. Run SR-17018 alongside morphine and, if possible, TRV130. Morphine grounds the assay in the established narrow-window reference; TRV130 gives a clinically-validated intermediate-bias point. Avoid drawing conclusions from SR-17018 alone.
• You want to test the signaling-bias framework itself. Include a compound whose bias and efficacy predictions diverge — a low-efficacy balanced agonist like buprenorphine, or a high-efficacy biased agonist if one becomes available — so the bias and intrinsic-efficacy hypotheses make different predictions in your hands. This is how Gillis 2020 designed its experiments.^[3]
• You want to study tolerance and chronic dosing. SR-17018 is the standout compound for this work because of the Grim 2020 morphine-reversal result. Compare against morphine at minimum; PZM21 is useful if you want to test whether the tolerance-reversal property is specific to SR-17018 or generalizes to other biased ligands.^[2]

Whichever design you choose, pre-register your bias calculation methodology and your respiratory-depression endpoint. Both of those choices have, in the published literature, been the specific places where replication failures originate.

Is SR-17018 a better opioid than oliceridine? It is more biased in published head-to-head cell-based characterizations and has a wider therapeutic window in the Schmid 2017 rodent comparison.^[1]It is not a clinical drug, so “better” in a practical sense is not the right question. Oliceridine has been run through human pharmacology and has a known clinical side-effect profile; SR-17018 has not.

Are SR-17018 and PZM21 the same kind of molecule? No. The two were discovered by different approaches — SR-17018 from a medicinal-chemistry series at Scripps, PZM21 from a structure-based virtual screen at UCSF — and they are distinct chemical entities. Both are biased MOR agonists in their original characterizations, but SR-17018 shows a substantially higher bias factor, and PZM21's claimed advantages have been partly overturned by follow-up work.^[6]

Does morphine have a bias factor of zero? By convention, DAMGO is often set to zero and morphine is measured relative to it. In most MOR datasets, morphine is very close to DAMGO — slightly G-protein-leaning in some comparisons, slightly arrestin-leaning in others, depending on the assay. It is the archetypal balanced reference in this literature.

Is SR-17018 an alternative to morphine for pain? No. It has not been studied in humans for any indication, is not approved for any clinical use, and is supplied strictly for research.