Choosing a reconstitution solvent looks like a one-line decision and usually is not. The same lyophilized solid can go into solution cleanly in one diluent, form a persistent haze in another, and dissolve fully in a third only to drop out four days later. What decides which happens is the sequence's charge state at the pH you put it in, how much hydrophobic surface it presents, whether the solvent contains anything besides water, and what the solid went through during freeze-drying. This page works through each and is explicit about which standard bench rules are measurements and which are convention.

Three different failures get called "it didn't dissolve"

Separate them before choosing a solvent, because they respond to different interventions. Slow dissolution: the material would reach target given time and the right conditions, but the cake is still sitting there. Saturation: equilibrium solubility in that solvent at that pH is below your target, and waiting does not fix it. Colloidal instability: the peptide dissolves, then self-associates into aggregates that may or may not be reversible.

A study of recombinant human tissue factor pathway inhibitor drew a version of this line usefully. The authors distinguished "insolubility precipitates" — those arising from solute saturation during buffer exchange, which redissolved under native solvent conditions — from "instability precipitates" seen after elevated temperature or extended storage, which were insoluble aggregate and required strongly denaturing conditions to redissolve [1]. That is a practical diagnostic: if material that fell out goes back into solution when you restore the original conditions, you were at saturation; if it does not, you have an aggregate, and that is a different problem from using a bigger volume.

Net charge, the isoelectric point, and where the usual rule breaks

The isoelectric point is the pH at which a molecule's net charge is zero. The bench heuristic that follows is that a peptide is least soluble near its pI, because charge-charge repulsion between molecules is at a minimum there.

Net charge can be estimated from sequence with arithmetic worth showing. Take a peptide with two lysines, one arginine, three glutamates, a free N-terminus and a free C-terminus. At neutral pH, using the conventional assumption that side-chain amines and guanidinium groups are protonated and carboxylates are deprotonated: lysine contributes +1 each (+2), arginine +1, the N-terminal amine +1, for +4; the three glutamate side chains contribute -1 each (-3) and the C-terminal carboxylate -1, for -4. Net charge is approximately zero, which places the pI near neutrality and predicts that plain water at ambient pH is close to the worst place to put it. Moving a pH unit in either direction shifts enough groups to take net charge off zero, which is the whole rationale for reaching for dilute acid or dilute base instead.

The heuristic is useful. It is not a law, and the measured exceptions get less credit than they deserve. In a systematic study that measured second osmotic virial coefficients for ovalbumin and catalase and aggregate and crystal solubilities for ovalbumin, beta-lactoglobulin A and B, ribonuclease A and lysozyme, aggregate solubility during salting-out decreased with decreasing pH for both the acidic and the basic proteins, and — explicitly against what the authors describe as common belief — neither aggregate nor crystal solubility had a minimum at the isoelectric point for the proteins examined [2]. The tissue factor pathway inhibitor work found an inverted bell-shaped pH-solubility profile with a broad minimum spanning pH 5 to 10, whose midpoint sat two to three pH units away from that protein's isoelectric point [1]. Both results are on proteins rather than short peptides, and neither author group generalized beyond the molecules they measured. The honest reading is that pI predicts the direction to move, not the exact pH of the minimum.

Net charge does track self-association in peptides where it has been quantified. For the model ionic-complementary peptide RADA16-I, size exclusion chromatography combined with simulation found a net charge of about +0.75 to be the condition promoting regular fibrils; lower net charges favored amorphous precipitates and higher net charges destabilized fibrils and promoted reversible monomer dissociation. The authors concluded that global net charge is a key property correlating well with fibril stability, while noting that conformation and the distribution of surface charge also contribute [3]. That last clause is the caveat to carry: net charge is the cheapest sequence-level predictor available, and it is not sufficient.

pH-driven association can be fully reversible. An anionic dermaseptin from frog skin was found to self-assemble into amyloid-like fibrils that disassemble cooperatively above pH 5, a transition the authors attributed to deprotonation of side-chain and terminal carboxyl groups producing intermolecular electrostatic repulsion; within the pH 5 to 6.5 transition interval, granular aggregates 10 to 500 nm in size formed instead [4]. For a peptide behaving this way, a cloudy solution is a pH problem, not a lost vial.

Hydrophobicity and why some sequences need a co-solvent

Charge explains the peptides that fail near their pI. It does not explain the ones that resist water at any pH. Long stretches of aliphatic and aromatic residues, lipidation, and amphipathic folds with a contiguous nonpolar face all present hydrophobic surface that water will not solvate favorably at useful concentrations. For these, the intervention is a co-solvent: a small fraction of an organic that solvates the nonpolar surface well enough to get the material into a homogeneous stock, which is then diluted into aqueous buffer.

Co-solvents are not conformationally inert, and this is the part most often skipped. Alcohols shift peptide secondary structure. Circular dichroism work on the S-peptide of ribonuclease A, an S-peptide analog, and a thermolysin fragment showed that trifluoroethanol, 1-propanol, 2-propanol and hexafluoro-2-propanol all enhanced helix content, with a plateauing dependence on co-solvent concentration; hexafluoro-2-propanol was the stronger helix enhancer in all three cases, but the relative effectiveness varied with the peptide sequence, and helix formation was pH-dependent in two of the three peptides [5]. So the co-solvent that rescues your solubility may also be holding the peptide in a conformation it does not adopt in the final aqueous system — which matters if the downstream measurement is conformation-sensitive.

Dimethyl sulfoxide deserves the same scrutiny. It is the default stock solvent in compound screening, where assay concentrations are typically 0.1% to 5% (v/v), corresponding to roughly 14 to 715 mM [6]. That is not a trace additive. In a study of two protein systems — the extracellular soluble domain of the human growth hormone receptor and a phosphatase domain — low amounts of DMSO destabilized both proteins investigated, altered the apparent binding property of one, and influenced the ionization process in electrospray mass spectrometry [6]. The authors' point was that standardizing on DMSO does not eliminate protein-specific effects; that argument transfers to peptides directly.

What the literature does not offer is a validated rule mapping sequence to co-solvent. Supplier solubility tables built on charge and hydrophobic-residue heuristics are useful starting points, not measured data for your sequence.

Bacteriostatic water versus sterile water

Sterile water contains nothing but water. Sterile water contains nothing but water. Bacteriostatic water contains benzyl alcohol as an antimicrobial preservative, which is what permits a vial to be entered more than once; the bacteriostatic amount used in the formulation literature is 0.9% w/v [8]. Worth converting: benzyl alcohol is C7H8O, so its molar mass is 7(12.011) + 8(1.008) + 15.999 = 84.08 + 8.06 + 16.00 = 108.14 g/mol. At 0.9% w/v — 0.9 g per 100 mL, or 9 g/L — the concentration is 9 ÷ 108.14 = 0.083 mol/L, about 83 mM. For scale: that is the same order of magnitude as the DMSO concentrations routinely present in screening assays, which [6] reports as 14 to 715 mM. The comparison is only a scale check — nothing about DMSO's behavior transfers to benzyl alcohol — but it makes the point that 83 mM is not a trace additive.

Benzyl alcohol is a common and reasonable choice: a review of antimicrobial preservative use found phenol and benzyl alcohol to be the two most frequently used preservatives in peptide and protein products, and identified the preservative's effect on the active ingredient as a key selection criterion alongside antimicrobial functionality. The same review noted that circular dichroism, fluorescence and differential scanning calorimetry are the techniques commonly used to evaluate a preservative's impact on conformational stability [7] — the interaction is treated as real enough to measure routinely.

The measured interactions run in one direction. Reconstituting lyophilized recombinant human interleukin-1 receptor antagonist with 0.9% w/v benzyl alcohol in water produced a greater degree of aggregation than reconstituting with water, and the extent of that aggregation was strongly modulated by how well native secondary structure had been retained through freeze-drying [8]. In recombinant human granulocyte colony stimulating factor, 0.9% benzyl alcohol accelerated aggregation at pH 7.0, an effect much greater at 37 °C than at 25 °C and partially counteracted by 1.0 M sucrose; at pH 3.5 it did not induce aggregation, although it perturbed tertiary structure at both pH values without detectably altering secondary structure. The authors postulated that the absence of aggregation at pH 3.5 reflected unfavorable energetics of intermolecular interaction rather than an absence of structural perturbation [9]. For recombinant human interferon-gamma, benzyl alcohol at 0.9% produced near-UV circular dichroism changes accompanied by high molecular weight aggregate formation, and the rate of that loss was accelerated at lower protein concentration and by increasing buffer or preservative concentration; the authors concluded the interaction is formulation-dependent, with protein concentration, buffer species and buffer concentration all contributing [10].

Two of these findings are counterintuitive enough to state plainly. The dilute stock is not automatically the safe one — in the interferon-gamma system, dilution made it worse [10]. And in the interleukin-1 receptor antagonist work the risk showed up at the moment of reconstitution rather than accumulating afterward: benzyl alcohol did not accelerate aggregation during subsequent room-temperature storage of the reconstituted samples [8].

None of this argues against bacteriostatic water. It argues that "preserved or unpreserved" is a formulation decision with a measurable effect on the molecule, not only a sterility decision. Where a stock will be entered repeatedly over days, the preservative is doing real work. Where a stock will be aliquoted and frozen immediately, or where benzyl alcohol would interfere with the downstream readout, unpreserved water removes a variable.

How solvent choice feeds back into stability

The solvent sets pH, and pH sets the dominant degradation chemistry. Deamidation of asparagine is the clearest worked example. In the model hexapeptide Val-Tyr-Pro-Asn-Gly-Ala, degradation depended markedly on pH, temperature and buffer composition: between pH 5 and 12 deamidation proceeded exclusively through a cyclic imide intermediate yielding both the aspartyl and isoaspartyl products, with buffer catalysis observed between pH 7 and 11, while at acidic pH the pathway was direct hydrolysis of the side-chain amide, giving only the aspartyl product [11]. Two consequences follow for solvent choice. Buffer identity is not neutral — the buffer species itself can catalyze the reaction in the neutral-to-alkaline range. And the pH that best solubilizes a sequence and the pH that best preserves it need not coincide; where they conflict, make that trade-off deliberately.

Freezing introduces a stress unrelated to the solvent's bulk composition. The ice-water interface is itself a site of protein destabilization; a review surveys the proposed mechanisms of ice-induced denaturation and the strategies that may minimize the resulting loss of activity, while presenting the area as not yet fully understood [12]. Carry that framing to the freezer: freeze-thaw is a known stress whose mechanism is still argued about.

Lyophilization can leave a peptide more aggregation-prone than it started, and this changes what reconstitution has to survive. Lyophilized formulations of the peptide teriparatide, PTH(1-34), were stable for months as dried solids, but physical stability after reconstitution varied considerably with peptide concentration and storage temperature, with precipitation seen within two to four weeks in some samples — while equivalent samples that had never been lyophilized showed no precipitation in the liquid state for as long as twelve weeks. The authors concluded that the peptide appears to adopt a higher-order structure perturbed by the combined stresses of freezing and drying, and that lyophilization can increase aggregation propensity even in a peptide rather than only in a protein [13]. The bench consequence: a reconstituted stock is not equivalent to a never-dried solution of the same peptide, and its behavior over weeks cannot be inferred from the dry material's shelf stability.

An order of operations

The following is standard bench practice rather than a protocol validated in the primary literature, and is offered as such.

  • Calculate net charge from the sequence at the pH you intend to use, and note the pI. Use it as a starting direction — dilute acid for a basic sequence, dilute base for an acidic one — recognizing that this is convention rather than a tested rule, and that the pI does not locate the solubility minimum [1, 2].
  • Test on a few milligrams before committing a vial.
  • Start with the simplest solvent that could work and escalate only when it does not: water, then pH-adjusted aqueous, then aqueous with a minimum fraction of co-solvent.
  • If a co-solvent is required, keep the fraction as low as will hold the material, and check that it does not interfere with the downstream measurement [5, 6].
  • Record what you actually did — solvent, lot, volume, date — on the vial, not only in the notebook.
  • Re-quantify after any filtration or centrifugation. What is in the supernatant is what dissolved; the nominal concentration is an assumption until measured.

Note also that nominal concentration and peptide concentration differ. A 10 mg vial brought up in 2 mL gives 5 mg/mL of solid, but if peptide content is 80% the actual peptide concentration is 0.80 × 10 mg = 8 mg in 2 mL, or 4 mg/mL. Solubility limits apply to the peptide, not to the solid.

What is not settled

Most of what is written above about proteins was measured on proteins. Short synthetic peptides are underrepresented in this literature, and extrapolating from a folded 150-residue protein to a 10-residue peptide is an assumption, not a result. The pI-minimum rule has documented exceptions in the systems where it was tested carefully [1, 2], and no general account of when it holds has been published. Sequence-based prediction remains approximate: net charge correlates with self-association where it has been quantified, but conformation and surface charge distribution contribute in ways that resist a single number [3]. Side-chain pKa values shift with local environment, so the net-charge arithmetic above is a first approximation, not a measurement. And benzyl alcohol's interaction has not been characterized for the overwhelming majority of sequences. Preserved peptide products exist — [7] reports benzyl alcohol as one of the two most common preservatives in peptide and protein products alike — but the published characterizations of what the preservative does to the molecule are all on proteins that happened to be formulated in repeatedly-entered vials [8, 9, 10]. Where it matters, measure it on the material in hand.