Isoelectric Point and Peptide Solubility: A Correction

The claim and the number underneath it

One sentence carries the entire circulating account of why tesamorelin gels on reconstitution: the isoelectric point sits near neutral pH, a near-neutral diluent therefore leaves the molecule with no net charge, repulsion between molecules collapses, and the peptide associates into a gel. The chain is coherent. Every link after the first depends on the first, and the first is a number. When the ledger entry for that compound went looking for the number, it found no measurement anywhere in the indexed literature, only supplier pages repeating the assertion without attribution. The ledger therefore computed a value instead, from the verified sequence, and published the computation as a computation.

That computed value came out near pH 11.3, which is not near neutral under any reading. The calculation below is re-run from scratch with the method stated, and the value moves with the pK set chosen. The mechanism the original claim invoked is real; the number attached to it was not. Peptide gelling has been measured in several distinct systems, and proximity to the isoelectric point is one of the routes documented in those systems.

Recomputing the isoelectric point

The sequence used is the mature somatoliberin chain, residues 32 to 75 of UniProt P01286, which reads YADAIFTNSYRKVLGQLSARKLLQDIMSRQQGESNQERGARARL and is annotated in that record as terminating in leucine amide. Tesamorelin adds a trans-3-hexenoyl group at the N-terminus. Both termini are therefore blocked, the amide removing the free carboxylate and the acyl group the free amine. Counting ionisable side chains across the 44 residues gives six arginines, two lysines, two aspartates, two glutamates and two tyrosines, with no histidine and no cysteine. The PubChem formula for CID 16137828, C221H366N72O67S, carries a single sulfur atom, consistent with one methionine and no cysteine.

Net charge was then calculated as a Henderson-Hasselbalch sum over those groups and the root located by bisection, which is the same charge equation the Expasy Compute pI/Mw tool solves, with the terminal terms removed to reflect the blocking. Using the Bjellqvist pK set that tool documents, in which Asp is 4.05, Glu 4.45, Cys 9.0, Tyr 10.0, His 5.98, Lys 10.0 and Arg 12.0, the isoelectric point falls at pH 10.90. Substituting the EMBOSS set, with Asp 3.9, Glu 4.1, Tyr 10.1, Lys 10.8 and Arg 12.5, moves it to 11.43. The ledger figure of approximately 11.3 sits inside that interval.

One result does not move with the parameters. Across both pK sets the net charge holds at roughly plus four from pH 5 through pH 8, reading plus 4.00 at pH 7.0 in each and plus 4.64 and plus 4.37 respectively at pH 5. That plateau is structural rather than parametric: eight basic side chains against four acidic ones, no histidine to titrate through the neutral range, and no free termini to contribute. Treating the termini as free instead of blocked drops the computed value to 10.40 on the Bjellqvist set and 10.91 on EMBOSS, and still leaves nothing near neutral.

A computed isoelectric point is not a measured one

A computed value and a measured value carry different error bars, and one published benchmark quantifies the gap. Kozlowski's 2016 benchmark of the Isoelectric Point Calculator compared fifteen published pK sets and two machine-learning predictors against experimentally determined values, and reported an average error for the best-performing sets of 0.87 pH units for proteins and 0.25 pH units for peptides. Errors of that magnitude are wider than the decimal places reported, large enough that a computed value has to be read as an estimate carrying a stated method, which is how the ledger publishes it and how this note repeats it.

Shaw and colleagues made the same point directly while engineering charge variants of ribonuclease Sa in vitro. Wild-type RNase Sa has a measured pI of 3.5 and contains no lysine residues; replacing surface aspartate and glutamate residues with lysine produced a three-substitution variant at pI 6.4 and a five-substitution variant at pI 10.2. That 2001 paper states explicitly that pI values estimated from model-compound pK data can be in error by more than one pH unit, and suggests how the estimation might be improved. A computed pI places a molecule approximately on the pH axis, within the error the benchmark above quantifies.

Net charge and the solubility minimum

Net charge on a peptide is a continuous function of pH rather than a fixed property. Each ionisable group is protonated or deprotonated in a ratio set by its dissociation constant and the surrounding pH, and the molecular total is the sum over all such groups. At the isoelectric point that sum passes through zero. Molecules carrying no net charge do not repel one another electrostatically, so the attractive terms, hydrophobic contact and dipolar and dispersion interactions, operate without opposition and association becomes correspondingly easier. Away from the isoelectric point, like-charged molecules repel, and that repulsion is what holds a concentrated solution together.

The empirical form of this is old and directly measured. Shaw and colleagues found, across RNase Sa and its two engineered variants, that the pH of minimum solubility varied with the pI of the protein while the pH of maximum activity and the pH of maximum stability did not, isolating charge as the variable the solubility minimum tracked. Trevino and colleagues subsequently measured solubility of RNase Sa variants in ammonium sulfate at high positive net charge, at low net charge and at high negative net charge, resolving the contribution of individual residue types on that same scale.

The original claim was reaching for a mechanism that exists. Its failure is arithmetic rather than physical, and applied to this molecule the mechanism runs backwards. A diluent near neutral pH leaves the peptide about four charges from neutrality and several pH units below its computed isoelectric point, which is the condition the same account elsewhere describes as favouring dissolution. Whatever produces the reported gelling, a collapse of intermolecular repulsion at neutral pH is not a candidate the sequence supports.

Beta-sheet aggregation is a separate route

Precipitation near the isoelectric point is not the only way a peptide solution turns cloudy or sets. Peptides also convert from soluble monomer into ordered, hydrogen-bonded beta-sheet assemblies, and the resulting fibrils entangle into gels. Glucagon, a 29-residue peptide hormone, is the system in which this behaviour has been documented longest, from 1969 onwards. Beaven, Gratzer and Davies described the formation and structure of gels and fibrils from glucagon in 1969. The record carries no abstract in the index; its MeSH indexing lists temperature, viscosity, sodium chloride and hydrochloric acid, and the full text was not reachable for this note, so what the paper manipulated as against what it measured cannot be stated here.

Pedersen and colleagues revisited the same peptide in 2006 with a time-resolved fluorescence assay in vitro at pH 2.5. Fibrils formed under different salt concentrations, glucagon concentrations and temperatures differed in kinetics, morphology, thioflavin T staining and infrared and circular dichroism spectra, and preformed fibrils used as seeds imprinted their properties on subsequent generations in a prion-like manner. The authors concluded that fibril structure was kinetically controlled by solvent conditions and by seeding rather than settled by a global energy minimum.

The equivalent process for insulin was characterised in vitro by Nielsen and colleagues, who reported that insulin concentration, agitation, pH, ionic strength, anion identity and seeding all altered fibrillation kinetics through a combination of hydrophobic and electrostatic interactions. That work recorded shorter lag times and faster fibril growth at acidic pH than at neutral pH, so fibrillation is not pH-indifferent; the glucagon work above resolves nothing on that axis, having been run at pH 2.5 throughout. What both papers establish instead is kinetic control. Fibrillation has a lag phase, it is accelerated by material already present, and it responds to the handling history of the sample rather than to a single equilibrium property of the molecule.

Agitation, interfaces and freezing

Mechanical history accounts for a further set of cases, and it is the variable the regulatory labelling addresses when it specifies mixing. Sluzky and colleagues examined insulin aggregation in aqueous solution in 1991 and reported that agitation rate, interfacial interactions and insulin concentration all governed the overall aggregation rate, with hydrophobic surfaces implicated directly. FDA prescribing information for tesamorelin covers two marketed presentations. Both address mixing by prohibiting shaking, while specifying different diluents and different mixing motions, and the labelling records that the two presentations are not substitutable. Those provisions constrain the air-liquid interface rather than the isoelectric point.

Koepf and colleagues took the liquid-air interface apart in 2018 using two monoclonal antibodies and a compression trough. The adsorbed protein formed a highly compressible film with an inhomogeneous distribution across the interface, and repeated compression and decompression produced considerable hysteresis together with significantly elevated numbers of particles. Infrared reflection-absorption spectroscopy indicated no considerable change in secondary structure relative to solution, meaning the antibody remained native-like while adsorbed. Particle formation on that evidence followed from interfacial packing and film collapse rather than from unfolding.

Freezing supplies an interface of its own. Chang, Kendrick and Carpenter tested several proteins in 1996 and found a correlation of r equals 0.99 between a protein's tendency to denature on freezing and its tendency to denature at a surface, attributing freeze-thaw damage primarily to the increase in ice-water interfacial area, with small amounts of surfactant protecting against both. Kueltzo and colleagues later put an IgG2 monoclonal antibody through freeze-thawing and reported that aggregation rose as pH fell, was most prevalent at pH 3 and 4, and varied sharply with container material.

What the record still does not contain

The searches behind the original ledger entry were re-run for this note, and one figure has changed. A PubMed title and abstract search for tesamorelin returned 94 records on 17 August 2026. Intersecting that set against isoelectric point, solubility, aggregation, formulation, gel, reconstitution and diluent returned two records on the same date, rather than the zero logged previously. Neither characterises the molecule in solution: one is a 2020 synthetic-chemistry paper using tesamorelin to demonstrate a peptide ligation method, the other a 2026 narrative review of injectable peptides that names the compound among growth hormone axis secretagogues. The finding stands with the count corrected.

No measured isoelectric point for this compound appears in the indexed literature. Nor does a solubility-versus-pH curve, an aggregation kinetics study, or a forced-degradation study. Every mechanism described above was measured on some other molecule, a bacterial ribonuclease, insulin, glucagon, or a monoclonal antibody, and none of those is a 44-residue acylated GHRH analogue. The mechanisms transfer as physics and not as numbers. The near-neutral pI claim stays logged as untraceable, now with a computed alternative beside it and the pK-set dependence of that computation stated in the open.