MID-35 Evidence Review: What Is Measured, What Is Not

MID-35 is a 16-residue, C-terminally amidated, all-D retro-inverso peptide developed from a myostatin-prodomain research programme. It has produced repeatable local muscle-weight effects in mice, including one cancer-cachexia model and experiments in young, adult, and aged animals. It also resists two isolated digestive proteases.

The evidence ends there sooner than many summaries admit. No study has measured MID-35 concentration in plasma, blood, urine, or any distant tissue. No study has reported its half-life, clearance, systemic bioavailability, or biodistribution. No human study, formal toxicology programme, immunogenicity study, independent laboratory replication, or direct structural measurement of a MID-35 to myostatin complex has been published. No study has measured contractile force or endurance in a MID-35-treated animal.

That combination makes MID-35 scientifically interesting and easy to overstate. The local mouse findings are real. Systemic exposure and human translation remain unknown. Kimera supplies MID-35 as a research reagent only.

This revision carries a search cutoff of 6 October 2026. Four peer-reviewed experimental papers name MID-35 and report original mouse work: Takayama 2022, Hanada 2022, Michiue 2023, and Morito 2026. Later papers that reprint the 2022 reporter values are not additional independent measurements. Three things have been added since the previous version of this page: the lineage’s own published systemic experiment, now read at figure level; the measured cost side of myostatin-pathway blockade, which this page previously omitted; and three corrections to numbers this page itself printed.

Identity and design

Takayama and colleagues introduced the name MID-35 in 2022. The published sequence is a 16-mer composed entirely of D-configured residues, including two D-cyclohexylglycine residues, with a C-terminal amide. One compact notation is lrxkrwirxkiwriyw-amide, where lower-case letters are D residues and x is D-cyclohexylglycine.

The residue-level sequence printed for patent Synthesis Example 23 is H-D-Leu-D-Arg-D-Chg-D-Lys-D-Arg-D-Trp-D-Ile-D-Arg-D-Chg-D-Lys-D-Ile-D-Trp-D-Arg-D-Ile-D-Tyr-D-Trp-NH2. The same printed sequence appears as riDM-35, SEQ ID NO: 22, in JP7740715B2, granted 17 September 2025, and in WO2022009698A1, published 13 January 2022. The grant page does not print the name MID-35. The defensible statement is that the sequence the papers name MID-35 matches riDM-35 / SEQ ID NO: 22. A patent is a dated public claim about what was asserted and when. It is not a second independent biology paper, and examination is not validation.

Patent Synthesis Example 23 reports high-resolution mass spectrometry calculated 2350.4657 and found 2350.4646. Claim 1 covers a 15 to 17 residue all-D formula with allowed substitutions. Claim 4 names SEQ ID NOs 20 to 23 and 25. Those analogues are not MID-35. Matching one sequence in a claim set does not make every claimed analogue the named compound.

That sequence crosswalk supports identity at the documentary level. It does not independently verify the contents of any commercial vial. No published study has sequence-confirmed a marketed lot against the 2022 research material.

Mass, registry identifiers and provenance

Sequence-derived average molecular weight is 2350.99 Da. Charge and basic-site counts that appear in some summaries are calculations from the sequence, not lot measurements. No assigned CAS number is held in the primary papers or the patent family. Catalogue SKUs are inventory labels, not identity measurements.

Michiue 2023 prints about 2.100 kDa in one methods sentence while its introduction and the sequence mass support about 2.35 kDa. Use the sequence-derived mass. Independent commercial listings print the same 2350.99 figure, which corroborates the arithmetic but is not a measurement of anything. The paper’s fluorescent tracer is also a different molecule from MID-35, treated later.

The originating institutions are Tokyo University of Pharmacy and Life Sciences with Fujita Health University. At least three outside sources, including one drug database, print a different originator. A database entry is not a provenance record.

Myostatin itself is a TGF-beta-family ligand that limits skeletal-muscle growth through ActRII and ALK4/5 signalling and Smad2/3. The natural inhibitor is the prodomain, not a receptor antagonist. MID-35 was built from that prodomain region. Native biology explains the design target. It does not prove that a 16-mer peptide will behave like a knockout mouse, an antibody, or a ligand trap.

How the 16-mer was built

MID-35 did not appear in one discovery step. It came from a decade of prodomain-derived peptide work in the same research network.

Jiang 2004 mapped myostatin-propeptide inhibitory activity to a broad region, residues 42 to 115 in that paper’s numbering, and could not separate binding from inhibition. Le 2022 later showed that protection of the prodomain alpha-1 helix correlates with latency across TGF-beta-family pro-complexes. Those papers explain why chemists hunted a short alpha-1-derived peptide. They do not study MID-35.

Takayama 2015 narrowed a mouse-prodomain peptide to a working 24-mer and measured SPR KD 35.9 nM. A 23-mer lacking one alanine measured KD 29.7 nM. Those binding constants belong to the parents, not to MID-35.

The parent 23-mer has three published IC50 values, not one

This page previously printed a single parent value and told readers to prefer it. That instruction was wrong, and the correction is more interesting than the original claim. The same laboratory measured the same 23-mer three times and published three different numbers.

Paper	Value for the parent 23-mer	Status
Takayama 2015, J Med Chem 58:1544	IC50 4.1 µM, as peptide 2	Measured. The founding paper. Its supporting information contains no IC50 at all
Takayama 2016, ChemMedChem 11:845	3.53 µM	Measured
Takayama 2017, ACS Med Chem Lett 8:751	3.56 ± 0.25 µM, as peptide 1	Measured, with the full four-parameter regression printed in supporting Table S1: IC50 3.562e-6 M, LogIC50 -5.448, HillSlope -3.721

The later literature flattened those three determinations into a single figure of 3.56 µM. In at least one case it attributed that figure to the paper that printed 4.1. Takayama 2020 cites 3.56 µM to its reference 6, which is Takayama 2015.

Cite 3.56 ± 0.25 µM to Takayama 2017, which measured it. Cite 4.1 µM to Takayama 2015, which measured it. Never present any one of the three as the IC50 of the 23-mer without naming the paper that measured it, because three independent determinations of one peptide by one group is itself part of the record.

The same pattern repeats one generation later. Takayama 2017 reports its hydrophobically modified analogue, peptide 3d, at IC50 0.32 ± 0.05 µM. Takayama 2019 reports the same molecule, which it identifies as “peptide 3d in ref 21”, at 0.29 ± 0.05 µM. Rentier 2019 reports it at 0.36 ± 0.17 µM. The error bars overlap and Rentier’s is much the widest, so no single value is contradicted. The point is the pattern, not any one number.

Chain shortening, and what the structural data can carry

Takayama 2019 shortened the series to MIPE-1686, a 16-mer with unnatural residues and one D-Trp. In that paper MIPE-1686 measured 0.13 ± 0.002 µM. In the 2022 MID-35 paper the same laboratory measured MIPE-1686 at 0.26 ± 0.04 µM. Same molecule, different assay. Cross-paper potency ranking is not allowed, and that includes ranking either value against MID-35’s 0.19 µM.

The lineage reports a consistent structural direction: alpha-helix propensity tracks with inactivity, and beta-sheet propensity tracks with potency and with protease resistance. A previous version of this page attached specific secondary-structure percentages to MIPE-1686. Those percentages have been removed for two reasons. Circular-dichroism secondary-structure percentages are deconvolution outputs rather than observations, so different fitting algorithms return different splits from one spectrum and the values are not comparable across papers. And the quantified breakdowns in this lineage are back-cited rather than measured in the paper that prints them, with no located primary source for the figure this page carried. The direction is well supported. The percentages are not publishable as structural composition.

Takayama 2019 also reported local mdx-mouse tibialis-anterior weight at 114 ± 6.6% of contralateral control and a grip increase after two 30 nmol intramuscular doses. Grip was 5.92 ± 0.82 versus 4.41 ± 0.31 kg/kg. Those results belong to MIPE-1686. They show that a local 16-mer in this lineage can move a functional readout in mdx mice. They do not transfer that grip result onto MID-35.

The 2022 paper constructed retro-inverso all-D versions of that scaffold. MID-35 contains two arginine substitutions beyond a direct retro-inverso copy. Retro-inversion reverses sequence order while replacing L residues with their D counterparts. This can preserve an approximate side-chain arrangement while changing backbone direction and protease recognition. It is a design strategy, not a guarantee of preserved structure, potency, or exposure.

The one systemic experiment in this lineage, and what it shows

A previous version of this page described the lineage’s intravenous experiment as unpublished. That was wrong. It is published, it is fully specified, and it is freely reusable. Ohsawa 2015 reports it as S4 Fig in PLOS ONE under CC BY. The compound is p29, an earlier 29-residue prodomain peptide, not MID-35.

Parameter	As printed
Route	Tail vein, intravenous
Dose	20 nmol, stated as the same amount as the intramuscular dose that worked
Schedule	Once a week, from 6 to 11 weeks of age
Animals	Wild-type and CAV3 P104L transgenic (caveolin 3-deficient) mice
n	5, as printed
Endpoints	Body weight gain, grip strength
Result	No effect on either

The authors offer two explanations in their own discussion, and they are mutually exclusive in practice:

The dosage of p29 for systemic administration could be insufficient to increase muscle mass. Alternatively, p29 peptide could be unstable against proteolysis when intravenously administered.

The paper does not choose between them, and it measured neither. No plasma concentration, no half-life and no biodistribution was determined in this experiment. It establishes that one systemic attempt at this dose and schedule produced no functional effect. It is not evidence that systemic delivery cannot work, and it is not evidence of a clearance mechanism.

Reading the figure rather than the sentence

Reading the figure adds two things. The experiment is stratified by sex into four arms, which neither the legend nor the main text states, so it is not determinable from the paper whether n = 5 is per arm, per genotype, or in total. That is a reporting gap and it should travel with any number taken from this figure. And on the functional endpoint the treated males track at or slightly below control males from week 7 onward in both genotypes. Error bars overlap throughout, so “no effect” remains the correct characterisation and no decrement may be claimed. But a reader given only the sentence would picture traces lying on top of each other, and that is not what the panel shows.

One further point matters for how the two research programmes relate. The author contributions on Ohsawa 2015 list both Takayama and Tsuchida. The prodomain-peptide work and the peptide-chemistry lineage are not parallel programmes that later converged. They are one collaboration from 2015 onward.

What was measured in cell-based assays

The 2022 originating paper used a Smad-responsive reporter in HEK293 cells. Ligand and peptide were preincubated for 20 minutes, then cells were treated for 4 hours. Each concentration was run in triplicate. Final ligand concentrations were 0.32 nM for myostatin, GDF-11, and activin A, and 0.1 nM for TGF-beta-1. MID-35 ranges were 0.025 to 2 µM against myostatin, 0.08 to 6 µM against GDF-11, and 0.025 to 18 µM against activin A and TGF-beta-1.

In that assay, MID-35 produced fitted IC50 values of:

0.19 µM against myostatin
0.63 µM against GDF-11
0.89 µM against activin A
1.6 µM against TGF-beta-1

On those fitted values alone, MID-35 was about 3.3-fold less potent against GDF-11, 4.7-fold less potent against activin A, and 8.4-fold less potent against TGF-beta-1 than against myostatin. Those ratios are assay-specific. They are not animal selectivity margins, safety margins, or receptor-level binding constants.

The panel also makes the label “selective myostatin inhibitor” too strong. GDF-11, activin A, and TGF-beta-1 all fell within roughly eightfold of the myostatin value. A precise description is that MID-35 showed a preference for myostatin in this assay while inhibiting all four tested ligands. Morito 2026 reprints the 0.19 / 0.63 / 0.89 µM set and describes dual activity. That is the originating group’s current interpretation, not a new potency experiment.

The prodomain region the series derives from behaves the same way. Ohsawa 2015 shows in its S3 Fig that the inhibitory core co-localises and co-immunoprecipitates with GDF11 as well as with myostatin. Cross-ligand activity is a property of the starting material, not an artefact introduced by chain shortening.

The activin A result remains disputed

Hanada and colleagues reported a different result later in 2022. Their HepG2 (SBE)4-luc experiment tested MID-35 at 1, 3, or 6 µM. Ligands were 10 ng/mL myostatin, 10 ng/mL GDF-11, 5 ng/mL activin A, and 5 ng/mL TGF-beta in the methods text. The main fixed-concentration experiment used 3 µM MID-35. Each transfection was triplicate and repeated at least twice.

Hanada reported inhibition of myostatin-, GDF-11-, and TGF-beta-induced activity, while activin signalling was unaffected. That conflicts with Takayama’s fitted activin A IC50 of 0.89 µM.

The two papers cannot be treated as equivalent experiments. Takayama fitted a concentration-response curve using 0.32 nM activin A. Hanada used a fixed 3 µM MID-35 condition, and its methods and figure caption disagree about whether the activin A concentration was 5 or 10 ng/mL and about the order of addition. Cell background and assay timing also differed. The same methods-versus-caption conflict already appears in Ojima 2020, a predecessor peptide-2 paper from the same laboratory, where the methods specify 5 ng/mL added after pretreatment and the Figure 1 caption specifies 10 ng/mL premixed with peptide.

The defensible conclusion is limited. Takayama measured activin A inhibition in one fitted reporter assay. Hanada reported no activin A effect in a different fixed-point assay with an internal protocol inconsistency. No independent laboratory has repeated the comparison side by side.

Hanada also measured immediate signalling. MID-35 only marginally reduced myostatin- and TGF-beta-mediated Smad2 phosphorylation at 3 µM, while the ALK4/5/7 kinase inhibitor SB-431542 abolished it. MID-35 did reduce myostatin-induced Smad2/3 nuclear accumulation and partly reversed effects on C2C12 myotube formation and marker expression. Reporter inhibition and complete suppression of the phosphorylation step are therefore not interchangeable claims.

A labelled species question, not a MID-35 prediction

Latres 2017 measured circulating GDF8 4 to 18-fold lower in monkey, rat, and human samples than in mice, while activin A was 3 to 4-fold higher in monkeys and humans than in mice and rats. Antibody blockade of myostatin alone increased mouse tibialis-anterior mass, while monkeys needed combination blockade of myostatin and activin A. Mass rose as much as about 49% with no change in specific force in any group.

That study used Regeneron antibodies, not MID-35. Every author was a company employee, and the company and five named authors hold patents on antibodies against both ligands, so a finding that primates need both blocked argues for the combination those patents cover. The conflict runs with the result and the result has not been independently reproduced. One untested hypothesis is that dual-ligand reporter inhibition might matter more in primates than in mice. No MID-35 primate, systemic, or human study exists to test it.

Direct binding and structure remain unmeasured

Earlier peptides in the lineage were tested for binding to myostatin, including 23- and 24-residue parents with dissociation constants near 30 to 36 nM. Those values do not belong to MID-35.

No published study has reported MID-35 binding by surface plasmon resonance, isothermal titration calorimetry, receptor competition, target occupancy, NMR, cryo-EM, crystallography, or hydrogen-deuterium exchange. No direct contact map shows where MID-35 binds mature myostatin, and no published measurement identifies which receptor interface it occupies. Docking models and residue lists from Asari’s work on earlier L-peptides describe those parents. Retro-inversion plus two arginine substitutions prevent a one-to-one transfer of that interface onto MID-35.

The lineage supports a reasonable working model. MID-35 likely retains enough of the prodomain-derived pharmacophore to reduce productive myostatin signalling. The exact binding surface and bound conformation remain hypotheses, and the experiment that would settle them is named at the end of this page.

Pathway design makes direct ActRIIB binding unlikely, because the series was built from a ligand-directed prodomain fragment. Unlikely is not measured. There is no MID-35 receptor-binding assay.

Close all-D relatives on the same scaffold produced circular-dichroism spectra with the opposite sign to L-peptide spectra from the same laboratory. That is qualitatively consistent with a mirror fold, and handedness is the part of the circular-dichroism record that carries. The published percentage fitting used L-protein references on un-inverted D spectra and is not usable as a quantitative MID-35 structure. MID-35 itself still lacks a bound or free structure.

Protease resistance

The strongest stability result belongs to two isolated-enzyme challenges in the 2022 paper. MID-35 at 50 µM was incubated for 400 minutes at 37 °C, n = 3. The reported intact fractions were 98.5 ± 7.8% with trypsin at 1 µg/mL and 97.0 ± 2.1% with alpha-chymotrypsin at 2 µg/mL. MID-35 peak area did not decrease significantly. The chromatograms showed no new product peak under those conditions.

MIPE-1686 measured 76.6 ± 9.7% and 72.0 ± 6.7% intact in the same comparison, and those decreases were significant. Trypsin produced a defined MIPE-1686 cleavage product by hydrolysis of the C-terminal Arg15 to Leu16 amide bond. A separate L-peptide positive control, SRIEWIKIQIWSKLRL-NH2, fell below 1% intact within 45 minutes in trypsin and 90 minutes in alpha-chymotrypsin, confirming that the enzymes were active.

This is strong evidence for resistance to those two isolated digestive proteases. It does not measure stability in whole blood, plasma, full serum, liver or kidney homogenate, intact tissue, or a living animal. It supplies no half-life or clearance number.

Why the earlier protease numbers cannot be merged with these

Takayama 2020 tested MIPE-1686 against three recombinant human proteases and reported 84.0 ± 3.4% remaining with aminopeptidase N, 90.1 ± 5.0% with chymotrypsin C, and 92.8 ± 2.4% with trypsin 3. Those figures belong to MIPE-1686, not to MID-35, and four features printed in that paper stop them being combined into one stability range with anything else.

The aminopeptidase N figure for MIPE-1686 is a 90-minute value. Every other figure in that table is at 45 minutes.
The three tables cover different peptide sets. One omits two peptides entirely, another omits two different ones. No row set is common to all three.
The three enzymes were used at 1, 2 and 0.1 µg/mL, so the percentages do not sit on a common scale of challenge severity.
The rows anyone would want to quote carry a non-specific-binding caveat and the others do not. In enzyme-free control incubations MIPE-1686’s peak area fell to 85 to 95% of its starting value through binding alone, while a comparator held at almost 100%. The authors read MIPE-1686’s reduction as non-specific binding rather than proteolysis and conclude it is “extremely stable against exopeptidases like APN”.

That last point cuts in MIPE-1686’s favour and against anyone quoting the number. An 84% figure sitting at or just below the binding floor is not a measure of enzymatic degradation. One more feature of the same table is worth stating: its lowest-scoring peptide is reported in the same paper as inactive, so the apparent trend from 22% to 84% does not track potency.

The same paper’s abstract states that MIPE-1686 “has a potential to serve as a long-lasting agent in vivo”. The paper contains no animal, no plasma sample and no half-life measurement. Protease resistance and elimination half-life are governed by different processes, and a peptide fully resistant to serum proteases can still clear quickly through hepatic uptake or renal filtration.

The complete published mouse record is local

Four papers report MID-35 efficacy experiments in mice. Every effective regimen placed the compound at or over the responding muscle. The studies do not share one control design.

Healthy young mice, one intramuscular dose

Takayama 2022 used eight-week-old male C57BL/6J mice. On day 0 they received one intramuscular injection of 30 nmol MID-35 (0.75 mM, 40 µL) into the left tibialis anterior and saline (40 µL) into the right. Muscles were collected on day 28.

The treated tibialis anterior weighed about 1.3 to 1.33 times the paired control. The paper reports MID-35 as stronger in vivo than MIPE-1686 despite similar reporter potency. That is consistent with greater local biochemical stability as a cause. It does not prove it, because no intact-peptide concentration was measured.

The study measured muscle weight after one local treatment. It did not include a functional endpoint, systemic dose, plasma sampling, tissue concentration time course, or biodistribution.

Cancer cachexia, three intramuscular doses

Hanada 2022 used a Lewis lung carcinoma cachexia model. Male C57BL/6J mice, 8 to 12 weeks old and 20 to 24 g, received 5 x 10^5 LLC cells subcutaneously. MID-35 was injected into both gastrocnemius muscles on days 4, 11, and 18 at 30 nmol per leg in 30 µL per leg. Main groups were n = 6. Because both legs were treated, this design cannot use an untreated contralateral muscle as an internal control.

Repeated local treatment partly preserved muscle-fibre area and increased grip strength relative to the cachexia control. At day 22, fibre cross-sectional area was 1817.2 ± 502.0 µm² in healthy controls, 1415.3 ± 504.7 µm² after cachexia plus PBS (a 22.1% reduction), and 1658.4 ± 424.3 µm² after cachexia plus MID-35 (an 8.2% reduction). Grip was significantly higher with MID-35 than PBS. The XML text does not print absolute grip values.

MID-35 did not significantly change survival, body weight, tumour growth, or heart weight in the first comparison. Unchanged heart weight is a cardiac-mass result. It is not a vascular or bleeding study.

The same paper paired MID-35 with oral anamorelin at 30 mg/kg on days 8 to 19. The combination produced the highest food-intake, water-intake, and grip measurements among the tested groups. Its survival comparison was p = 0.052, which did not meet the paper’s p < 0.05 threshold, so the study did not establish a statistically significant survival benefit. The combination arm showed a non-significant trend toward larger tumour volume, p = 0.09.

The combination did not show synergistic suppression of the measured atrophy-marker mRNAs. MID-35 alone most strongly reduced Smad2/3 nuclear localization and the expression of the measured atrophy-associated genes. These findings come from one tumour model and one cell line. They do not constitute a general cancer-safety result.

Iontophoretic delivery, three sessions

Michiue 2023 tested iontophoresis, a weak electrical current used to move material through the skin. Five-week-old male C57BL/6J mice, n = 5 per group, received 0.18 mg / 75 nmol MID-35 in 50 µL water under the anode. Current was 0.34 mA/cm² for one hour on days 0, 7, and 14. Tibialis anterior was collected on day 42. Reported purity was above 95% after reverse-phase HPLC.

Tibialis-anterior weight was 1.25-fold higher at day 42 and statistically significant. Gastrocnemius weight increased only slightly and not significantly. Mean fibre cross-sectional area did not differ from controls. Fractions of small 250 to 500 µm² fibres and large fibres above 2750 µm² tended to increase, but those distribution differences were not significant. Central nuclei appeared throughout treated tibialis-anterior sections.

The paper’s fluorescent tracking experiment did not label MID-35. It used FITC-epsilon-Acp-WYIEWIKIQIWSKLRL-NH2, a 17-residue all-L precursor analogue carrying FITC and aminocaproic acid. After one hour, fluorescence reached muscle-associated tissue at least 1000 µm below the skin. That shows the labelled comparator travelled into the treated region. It does not establish intact MID-35 concentration, absorbed dose, or systemic confinement.

RNA was measured 3, 6, and 24 hours after iontophoresis. Atrogin-1/Fbxo32 and MuRF-1/Trim63 rose at 3 hours and fell at 6 hours, and MID-35 enhanced those transient changes. That direction is opposite a simple expectation that blocking myostatin would immediately suppress atrophy markers. Morito 2026 later observed both genes decrease after direct intramuscular MID-35. Different route, dose, schedule, and sampling mean the two experiments should not be collapsed into one direction.

The authors compare 1.25-fold after three 75 nmol iontophoretic sessions with Takayama’s roughly 1.33-fold after one 30 nmol intramuscular injection. The regimens are not dose-equivalent, absorbed dose was never measured, and the studies were not head-to-head. Arithmetic comparison is not bioavailability.

Duration, age, dose, and sphingolipids

Morito 2026 provides the longest time course. Male C57BL/6JJmsSlc mice were young (8 weeks), adult (20 weeks), or aged (72 weeks). Main comparisons used n = 5, with MID-35 in the left tibialis anterior and saline in the right muscle of the same mouse. MID-35 was injected once in 40 µL on day 0. Most experiments used 2 nmol, with supporting experiments at 0.5 nmol.

In young mice, muscle weight first increased significantly at day 14, peaked in the measured series around day 28, and remained significantly elevated at day 84. The contralateral saline muscle did not show the same divergence. The lower 0.5 nmol level remained effective to day 84 in young mice. Adult mice showed significant hypertrophy after 2 nmol but not 0.5 nmol. Aged mice responded to 2 nmol.

Early molecular changes preceded the weight change. In young mice at day 3, filed graph-read estimates were Pax7 up about 2.3-fold, Myod1 up about 4-fold, Myog up about 7.3-fold, Trim63/MuRF1 down to about 30%, and Fbxo32/Atrogin-1 down to about 20%. Those transcript values are estimates read from published figures, not printed tables. By day 28, only Fbxo32 remained significantly reduced. By day 84, none of these measured expression differences remained significant. Centralized nuclei and Pax7 staining supported a regenerative or remodelling response, and the centralized-nucleus fractions this page previously quoted were graph reads that did not match the filed extraction, so only the direction is stated here.

In aged mice, tibialis-anterior weight, the fraction of large 70 to 80 µm fibres, and centralized nuclei increased. The fibre-size distribution differed from younger animals: 20 to 40 µm fibres increased and 50 to 60 µm fibres decreased. Pax7 and Myod1 responses were less than twofold, Myog did not increase, and the authors interpreted differentiation responsiveness as blunted with age.

Sphingolipids as an early associated signal

The same study measured sphingolipids in treated muscle. Sphingosine 1-phosphate and sphingosine rose at day 3 in young and adult mice. In young mice, S1P was lower in treated than control muscle at day 28, while sphingosine remained elevated. Aged mice showed a sphingosine increase without a significant S1P increase. By day 84, the measured lipid differences were absent while the weight effect persisted. Multiple ceramide species and sphingolipid-enzyme transcripts also shifted early. The authors state that none of the measured factors explained maintenance at day 84.

These are temporal associations. The experiment did not prove that S1P caused the muscle response, that S1P is required, or that the same pathway operates after systemic exposure. The authors infer that direct pharmacodynamic action was limited to less than seven days. That is an authors’ pharmacodynamic bound. It is not a peptide concentration, residence time, or pharmacokinetic half-life.

Duration of effect is not duration of exposure

The day-84 result is easy to misread. It shows persistence of a tissue outcome, not persistence of intact peptide.

One evidence-based hypothesis is that a short local pathway perturbation starts a regenerative process whose structural result lasts after direct peptide action has ended. Early transcriptional and lipid changes, followed by later weight gain and normalization of the measured early signals, fit that model. The model remains provisional because no study quantified intact MID-35 in muscle over time. A long local residence time has not been excluded. What would refute it is a concentration-time series in treated muscle showing intact peptide still present at day 84.

The cost side of the pathway, which this page previously omitted

Blocking myostatin reliably adds muscle mass. Whether the added mass performs is a separate question, and the primary literature on it is not flattering. This section is the counterweight, and it carries one hard boundary: not one of the studies below used MID-35. Five used a decoy receptor, a genetic knockout, or a fusion protein. One used propeptide overexpression. A transgene is not a drug, and results do not travel between molecule classes. They are included because they measure the endpoints MID-35 has never been tested on.

Mass does not buy proportional force in the genetic model

Giannesini 2013 measured myostatin-null mice non-invasively in vivo by proton MRI and phosphorus-31 MRS during maximal repeated isometric contractions. Body weight rose 38%, gastrocnemius volume rose 118%, and absolute force rose 34%. A 118% volume increase returning a 34% force increase is a large specific-force deficit. The same paper states that force-generation findings in this model have been studied mainly in vitro and “remain conflicting”, which is a primary-source admission rather than an outside criticism.

The fibre-type explanation is wrong, and the control experiment is the reason

The tidy story is that genetic myostatin loss shifts fibres toward a glycolytic phenotype, that the shift raises fatigability, and that fibre type therefore explains the cost. Mouisel 2014 ran the experiment that breaks it. Alongside knockout animals, that study overexpressed myostatin propeptide in adult mice, which “did not affect fiber-type distribution, while nonetheless causing increased muscle fatigability, diminished exercise capacity, and decreased Pparb/d and Pgc1a expression.”

Fatigability appeared with fibre-type distribution unchanged. The cost is metabolic rather than typological, and it runs through Pparb/d and Pgc1a. This matters for this compound class specifically, because Mouisel’s adult intervention was propeptide overexpression, and the whole lineage behind MID-35 is prodomain- and propeptide-derived. It is the closest mechanistic analogue in this evidence set, and it produced the cost.

Duration changes the answer

Two studies used the same molecule, the same dose and the same route, and reached apparently opposite conclusions.

	Béchir 2016	Relizani 2014
Agent	Soluble ActRIIB-Fc	Soluble ActRIIB-Fc
Dose and route	10 mg/kg subcutaneous, twice weekly	10 mg/kg subcutaneous, twice weekly
Duration	8 weeks	4 months
Fatigability	Unaffected	“Extreme muscle fatigability”
Mitochondria	ATP-producing capacity “dramatically reduced”	Porin reduced, oxidative phosphorylation deficient, capillarization reduced

At 8 weeks the metabolic lesion is already measurable while function still holds. At 4 months the function has given way. Béchir is not a reassurance; it is the early timepoint of Relizani’s result. Any page quoting the 8-week “fatigue resistance remained unaffected” line without the 4-month study is quoting an interim reading as a final one.

The study that went the other way, and what it traded

Boido 2020 fused a soluble activin receptor to agrin’s nLG3 domain through an IgG1 Fc. Young male mice on rotarod showed “a remarkably increased endurance”, significantly longer than the single-domain compounds. The trade is in the paper’s own title: the construct curbs the hypertrophic effect of myostatin inhibition. The endurance gain came with “only a moderate increase in body weights and wet muscle weights”, while plain ActR-Fc gave a highly significant weight increase. Dosing the two domains as a mixture did not reproduce the endurance effect. Two authors are at the company developing the construct, and that conflict runs with the finding.

Less mass, more endurance. It is the cleanest statement in this literature that the two are not the same axis.

The lineage’s own data on this axis

Ohsawa 2015 is the one place where a prodomain peptide was measured on force rather than weight. Specific tetanic force was significantly reduced in untreated caveolin 3-deficient muscle compared with wild type, and p29 injection increased muscle-specific force in both the dystrophic and the wild-type animals, n = 10. On the single endpoint that the knockout and decoy-receptor studies report as unchanged or worse, the prodomain peptide moved it upward in two genotypes.

The same paper’s fibre data complicate that. p29 increased the size and number of fast glycolytic type IIB fibres but not type I or type IIA, and the authors conclude that “p29 treatment changed the myofiber distribution to a faster glycolytic phenotype”, consistent with myostatin-null muscle. The fibre-size histograms show the treated distributions flattened and broadened rather than shifted, and the printed standard deviations roughly double with treatment in both genotypes: 2962.5 ± 476.3 to 3271.0 ± 823.7 µm² in wild type, and 1800.7 ± 224.9 to 2858.5 ± 428.5 µm² in the transgenic animals. Treated muscle therefore contains fibres inside the control range and fibres well beyond it. “Increased fibre size” is accurate and incomplete. Whether that heterogeneity is desirable is not addressed by the paper, and none of it is a MID-35 measurement.

Why none of this has translated in people

Mariot 2017 analysed human biopsies across five limb-girdle muscular dystrophy subtypes and proposed a reason: the myostatin pathway is already downregulated in diseased human muscle, so less signal remains to block. Its own summary of the clinical record is that significant improvements in muscle strength or physical function “have not been reached”, with the exception of two small open-label studies and one small randomised trial. The paper states that its data “may explain the poor clinical efficacy of anti-myostatin approaches in several of the clinical studies and the apparent contradictory results in mice”. It is published under CC BY and it is the single most useful contextual paper for anything written about this pathway.

The honest sentence for the whole section is that myostatin-pathway blockade reliably adds muscle mass, that the force and endurance consequences are mixed and intervention-specific, and that the longest pharmacological study in this set ends with its authors advising against the strategy in isolation.

The clinical field in 2026

MID-35 has no clinical data. The field it sits in changed twice in 2026, in opposite directions, and both changes bear on how this compound should be described.

An approval, and the claim it retires

The FDA approved apitegromab, marketed as ISEMBYLD, on 11 September 2026 for spinal muscular atrophy in patients aged 2 and over, for use alongside existing SMN-targeted therapy. It is an anti-latent-myostatin antibody and it is the first agent targeting myostatin to be approved for a muscle indication. Its prescribing information carries a warning about increased fracture risk, which was not anticipated from the earlier public record. Any statement that myostatin inhibition has never worked in humans was defensible through early 2026 and is not any more. That includes the narrower version of the claim: luspatercept, which acts in this receptor family, is approved for anaemia rather than for muscle, and it is no longer the only approval in the neighbourhood.

A programme with confirmed target engagement that failed anyway

In March 2026 the originator of emugrobart, an anti-latent-myostatin antibody, announced discontinuation of its phase 2/3 study in spinal muscular atrophy and its phase 2 study in facioscapulohumeral muscular dystrophy. The stated reason is the useful part. The announcement reports that the antibody “showed a favorable safety profile and target engagement by reducing mature myostatin” but that this “did not translate into the intended functional outcomes. Specifically, muscle growth and exploratory functional efficacy were neither consistent nor robust enough.”

Read together, the two events say something more precise than either alone. Suppressing myostatin is achievable in humans and has now produced one approval in one genetic disease on top of an existing disease-modifying therapy. Suppressing myostatin is also demonstrably not sufficient on its own, measured by the sponsor’s own target-engagement data against its own functional endpoints. Target engagement is not efficacy, and mass is not function. Neither of those principles is a criticism of MID-35, which has never been measured against a functional endpoint in any species at all.

Antibody, ligand-trap, and related programmes used different molecules, routes, and exposures from a locally injected 16-mer peptide. They do not measure MID-35. They do establish that a myostatin-pathway effect in mice, or a mass change in humans, does not automatically become a useful functional result. Receptor-side comparators in this pathway, including ACE-031, are collected in the peptides research library.

Study design limits

The published direction is consistent across four local mouse experiments, but the evidence base is narrow.

The studies come from an overlapping author network, and the prodomain and peptide-chemistry programmes have shared authors since 2015. No independent laboratory has replicated MID-35.
Group sizes were small, commonly five or six animals. Takayama 2015’s parent-peptide mdx work used n = 3.
Regimens, ages, models, and readouts differed.
Control structures differ and must not be collapsed into one design. Takayama 2022 and Morito used contralateral saline controls. Hanada treated both gastrocnemius muscles. Michiue used separate control groups. A paired contralateral design cannot detect an equal bilateral effect.
Functional testing was incomplete. Hanada reported a grip improvement in cachectic mice. The healthy-mouse and duration studies reported muscle mass without a direct force measurement. No study has measured specific force, contractile function, or endurance in a MID-35-treated animal.
No study verified a marketed lot against the research material used in the papers.
Purity, identity, and content are different measurements. HPLC purity reports the fraction of peptide-related material in the main peak and not how much peptide is in the vial. Mass spectrometry can support identity, and because L to D substitution is mass-silent, mass alone cannot confirm that a D-peptide was built correctly. None of those release tests is a pharmacokinetic measurement, and no validated bioanalytical method has been published for peptides of this class.

Kimera’s own position on that last point is that a certificate should be read off the certificate. Lot documentation with multi-lab third-party verification is published in the COA archive rather than summarised as a catalogue-level claim, and a specification is not a measurement no matter whose page it sits on.

The replicated local mouse muscle-weight effect deserves moderate confidence. Claims about systemic treatment, whole-body muscle effects, function, safety, or human benefit do not.

Errors that circulate around the named compound

Several published defects sit next to MID-35 and should be named rather than silently repaired. Three of them were in earlier versions of this page.

The parent 23-mer has three measured IC50 values, 4.1, 3.53 and 3.56 µM, from three papers by one group. The defect is not the existence of 3.56 µM, which was measured and printed with its full regression in 2017. The defect is Takayama 2020 attributing 3.56 µM to the 2015 paper, which prints 4.1 µM.
The capped analogue peptide 3d has three published values as well: 0.32, 0.29 and 0.36 µM. No paper acknowledges the others.
MIPE-1686’s 0.13 µM and 0.26 µM values come from different papers and must not be ranked against MID-35’s 0.19 µM.
Hanada’s methods and figure caption disagree on activin A concentration and order of addition, reproducing the same conflict in Ojima 2020.
Michiue prints about 2.100 kDa in one methods sentence. Sequence mass is about 2.35 kDa.
Michiue’s fluorescent 17-residue tracer is the all-L precursor analogue, not MID-35.
The secondary-structure percentages this page used to print for MIPE-1686 are back-cited in the paper that reports them and their primary source has not been located. They have been removed rather than repeated.

These are reasons to read primary pages and keep molecules, assays, and protocols separate. They are not a brief against the local mouse results.

The patent examples repeat the isolated trypsin and alpha-chymotrypsin challenge and an intramuscular cachexia grip experiment for riDM-35. That is the same programme restating the same work in a legal document, not an independent replication. Baig 2022 and related D-peptide analogues sit in a parallel optimization of the capped 22-mer lineage and are not MID-35. Follistatin-derived short peptides, FLRG and FSTL3 are different molecules, and the letter-swap names must stay separate. The transposition problem is old enough that the 1998 paper which named FLRG misspells it as FRLG in its own results.

What remains unknown

The following gaps are direct absences in the published record:

no MID-35 concentration in blood, plasma, serum, urine, or distant tissue
no Cmax, AUC, bioavailability, half-life, clearance, or mass-balance study
no systemic efficacy regimen for MID-35
no measurement of specific force, contractile function, fatigue resistance, or metabolic cost in a MID-35-treated animal
no human administration or clinical trial
no formal acute or repeat-dose toxicology programme
no immunogenicity study
no direct MID-35 binding constant or bound structure
no independent laboratory replication
no validated bioanalytical method published for peptides of this class

The no-concentration finding is not an empty search. Author, sequence, patent, and phrase queries through PubMed, Europe PMC, patent text, and the four primary papers all return local reporter, enzyme-tube, or intramuscular and iontophoretic work. Papers that reprint the 2022 IC50 table do not fill that gap. A commercial-listing sweep returns the same answer from the other direction: no vendor of this compound claims a half-life, a pharmacokinetic figure, or a duration of action for it.

What the chemical class permits, and why the gap is conspicuous

Small, protease-resistant peptides can still leave circulation quickly through distribution and renal filtration. Charge, protein binding, formulation, and tissue uptake can change that behaviour, and none of those factors has been measured for MID-35. The class evidence runs in a direction worth stating plainly, because it removes a common excuse.

An all-D peptide of comparable length, RD2, is reported with a terminal plasma half-life in mice of more than two days and high bioavailability after intraperitoneal, subcutaneous and oral administration. That is an abstract-level reading, not a full-text hold. The one retro-inverso peptide in the class whose disposition this review holds in full, studied as a radiolabelled conjugate in rats, cleared by a hepatobiliary route with a liver plateau reached within ten minutes. No published in vivo pharmacokinetic study exists for any 16-residue all-D retro-inverso peptide, so nothing in the literature sits at the intersection MID-35 occupies.

The consequence runs both ways and should be read both ways. Long systemic circulation is achievable for an all-D peptide of this size, so the absence of any MID-35 exposure measurement is more conspicuous rather than less. It is also not evidence that MID-35 has good systemic exposure. It removes one specific explanation, that D-peptides do not survive in blood, from the list of available reasons why every efficacy experiment in this lineage is local. Ohsawa 2015’s own alternative explanation, proteolytic instability, is coherent for p29 because p29 is an L-peptide. It is a weaker explanation for a retro-inverso all-D peptide.

What would settle the major questions

The highest-value next study would combine intact-parent pharmacokinetics with local pharmacodynamics. It would measure MID-35 and metabolites after both local and intravenous dosing in plasma, urine, kidney, liver, treated muscle, and distant muscle, which would separate local residence, systemic escape, clearance, and tissue exposure.

A second study should repeat full concentration-response curves against myostatin, GDF-11, activin A, and TGF-beta-1 under matched conditions, with direct binding and p-Smad measurements. That would address the activin A dispute and replace reporter-only selectivity ratios with a stronger mechanism record.

An independent, blinded mouse replication should include predeclared statistics, local and distant controls, muscle mass, histology, direct contractile function, and a fatigue or endurance protocol. The literature reviewed above shows that muscle weight can change without proving an improvement in force, and that the cost of blocking this pathway grows with the length of the experiment. A 28-day mass endpoint cannot see either.

A direct surface-plasmon-resonance or equivalent binding measurement, plus a contact map, would settle whether MID-35 occupies the parent peptide’s proposed interface or a different surface.

An independent sequence, purity, and reporter measurement on any marketed material would be required before applying the published MID-35 findings to a vial. The papers and the patent describe research synthesis. They do not analyse a commercial lot.

Bottom line

MID-35 is a documented retro-inverso all-D 16-mer that inhibits myostatin pathway reporters, resists two isolated digestive proteases, and increases locally treated muscle weight in several mouse experiments. The strongest functional result is improved grip in one cancer-cachexia model. The longest tissue result is a day-84 local weight increase after one intramuscular injection.

Its limits are equally clear. The compound is not strictly selective in the originating four-ligand assay. Its activin A activity is disputed. Its exact binding mode is unmeasured. Its durable tissue effect does not establish durable molecular exposure. Its own lineage contains one published systemic experiment, which produced no functional effect and measured no exposure. And the pathway it acts on carries a measured metabolic and endurance cost in other molecule classes that nobody has yet looked for in this one.

MID-35 is best described as a promising local mouse research tool with an unresolved exposure profile and an untested functional profile. Anything stronger runs ahead of the published data.

Primary references

Takayama K, Hitachi K, Okamoto H, et al. Development of Myostatin Inhibitory d-Peptides to Enhance the Potency, Increasing Skeletal Muscle Mass in Mice. ACS Med Chem Lett. 2022;13:492-498. doi:10.1021/acsmedchemlett.1c00705

Hanada K, Fukasawa K, Hinata H, et al. Combination Therapy with Anamorelin and a Myostatin Inhibitor Is Advantageous for Cancer Cachexia in a Mouse Model. Cancer Sci. 2022;113:3547-3557. doi:10.1111/cas.15491

Michiue K, Takayama K, Taniguchi A, Hayashi Y, Kogure K. Increasing Skeletal Muscle Mass in Mice by Non-Invasive Intramuscular Delivery of Myostatin Inhibitory Peptide by Iontophoresis. Pharmaceuticals. 2023;16:397. doi:10.3390/ph16030397

Morito K, Nishikawa N, Hitachi K, et al. Myostatin Inhibitory D-Peptides Induce Skeletal Muscle Hypertrophy along with Alteration of Bioactive Sphingolipid Metabolism. ACS Pharmacol Transl Sci. 2026;9:1544-1553. doi:10.1021/acsptsci.6c00124

Ohsawa Y, Okada T, Nishimatsu S, et al. The Inhibitory Core of the Myostatin Prodomain: Its Interaction with Both Type I and II Membrane Receptors, and Potential to Treat Muscle Atrophy. PLOS ONE. 2015;10:e0133713. doi:10.1371/journal.pone.0133713

Takayama K, Nakamura A, Rentier C, et al. Identification of the Minimum Peptide from Mouse Myostatin Prodomain for Human Myostatin Inhibition. J Med Chem. 2015;58:1544-1549. doi:10.1021/jm501170d

Takayama K, Noguchi Y, Aoki S, et al. Effect of N-Terminal Acylation on the Activity of Myostatin Inhibitory Peptides. ChemMedChem. 2016;11:845-849. doi:10.1002/cmdc.201500533

Takayama K, Rentier C, Asari T, et al. Development of Potent Myostatin Inhibitory Peptides through Hydrophobic Residue-Directed Structural Modification. ACS Med Chem Lett. 2017;8:751-756. doi:10.1021/acsmedchemlett.7b00168

Takayama K, Asari T, Saitoh M, et al. Chain-Shortened Myostatin Inhibitory Peptides Improve Grip Strength in Mice. ACS Med Chem Lett. 2019;10:985-990. doi:10.1021/acsmedchemlett.9b00174

Rentier C, Takayama K, Saitoh M, et al. Design and synthesis of potent myostatin inhibitory cyclic peptides. Bioorg Med Chem. 2019;27:1437-1443. doi:10.1016/j.bmc.2019.02.001

Takayama K, Odagiri M, Taguchi A, Taniguchi A, Hayashi Y. Enzymatic Stability of Myostatin Inhibitory 16-Mer Peptides. Chem Pharm Bull. 2020;68:512-515. doi:10.1248/cpb.c20-00158

Ojima C, Noguchi Y, Miyamoto T, et al. Peptide-2 from mouse myostatin precursor protein alleviates muscle wasting in cancer-associated cachexia. Cancer Sci. 2020;111:2954-2964. doi:10.1111/cas.14520

Jiang MS, Liang LF, Wang S, et al. Characterization and Identification of the Inhibitory Domain of GDF-8 Propeptide. Biochem Biophys Res Commun. 2004;315:525-531. doi:10.1016/j.bbrc.2004.01.085

Le VQ, Iacob RE, Tian Y, et al. TGF-beta Family Pro-complexes Pass through an Unstable State during Activation. J Mol Biol. 2022;434:167439. doi:10.1016/j.jmb.2021.167439

Latres E, Mastaitis J, Fury W, et al. Activin A More Prominently Regulates Muscle Mass in Primates than Does GDF8. Nat Commun. 2017;8:15153. doi:10.1038/ncomms15153

Giannesini B, Vilmen C, Amthor H, Bernard M, Bendahan D. Lack of myostatin impairs mechanical performance and ATP cost of contraction in exercising mouse gastrocnemius muscle in vivo. Am J Physiol Endocrinol Metab. 2013;305:E33-E40. doi:10.1152/ajpendo.00651.2012

Mouisel E, Relizani K, Mille-Hamard L, et al. Myostatin is a key mediator between energy metabolism and endurance capacity of skeletal muscle. Am J Physiol Regul Integr Comp Physiol. 2014;307:R444-R454. doi:10.1152/ajpregu.00377.2013

Relizani K, Mouisel E, Giannesini B, et al. Blockade of ActRIIB signaling triggers muscle fatigability and metabolic myopathy. Mol Ther. 2014;22:1423-1433. doi:10.1038/mt.2014.90

Béchir N, Pecchi E, Vilmen C, et al. ActRIIB blockade increases force-generating capacity and preserves energy supply in exercising mdx mouse muscle in vivo. FASEB J. 2016;30:3551-3562. doi:10.1096/fj.201600271RRR

Boido M, De Amicis E, Pallottini V, et al. Increasing agrin function antagonizes muscle atrophy and motor impairment in spinal muscular atrophy. PLOS ONE. 2020;15:e0228653. doi:10.1371/journal.pone.0228653

Mariot V, Joubert R, Hourdé C, et al. Downregulation of myostatin pathway in neuromuscular diseases may explain challenges of anti-myostatin therapeutic approaches. Nat Commun. 2017;8:1859. doi:10.1038/s41467-017-01486-4

Leithold LHE, Jiang N, Post J, et al. Pharmacokinetic Properties of a Novel D-Peptide Developed to be Therapeutically Active Against Toxic beta-Amyloid Oligomers. Pharm Res. 2015;33:328-336. doi:10.1007/s11095-015-1791-2

Peptides and complexes containing the same. JP7740715B2, granted 17 September 2025. SEQ ID NO: 22 / riDM-35.

Peptides and complexes containing the same. WO2022009698A1, published 13 January 2022.

Research use only

Kimera Chems supplies MID-35 strictly for laboratory research and analytical use. It is not a drug, food, cosmetic, or dietary supplement. It is not for human or veterinary use and must not be administered to humans or animals. Nothing on this page describes or implies a therapeutic use, and no statement here has been evaluated by the FDA. Sales are limited to qualified research professionals and institutions.