Knowledge

From Precursor to Insulin: What the Expression Route Decides

Insulin is not expressed directly as the finished two-chain molecule. It is expressed as a precursor, and everything downstream — how the material is captured, how many enzyme steps are needed, which impurities turn up in release testing — follows from which precursor and which host were chosen at the start.

Two production routes dominate the published work: expression in Escherichia coli , where the precursor accumulates inside the cell as insoluble inclusion bodies, and secretory expression in the yeast Pichia pastoris (syn. Komagataella pastoris ), where an insulin precursor is pushed out into the culture medium [5][2].

The bacterial route: fusion protein, inclusion bodies, conversion

Proinsulin is an intermediate between preproinsulin, the initial product of ribosomal translation, and insulin itself [5]. To be produced efficiently in bacteria it must be made as an N-terminally extended fusion protein, which then has to be converted to authentic proinsulin during purification [5]. That conversion step is not optional — it is a structural consequence of the expression strategy.

Proinsulin forms inclusion bodies during bacterial expression [5]. This can be exploited as a capture step: the dense insoluble particles are easy to separate from most soluble cell contents. But conventionally prepared inclusion bodies contain significant amounts of non-protein contaminants that interfere with subsequent chromatographic purification [5]. A reported method addressed this by altering the inclusion body preparation conditions so that the interfering contaminants are excluded while the proinsulin fusion protein is retained at high yield [5].

Two further problems in the bacterial route are worth naming because they recur in process development. The proinsulin molecules inside the inclusion bodies are incorrectly folded and likely cross-linked to one another, which makes it difficult to quantify how much proinsulin was actually expressed [5]. The reported answer was procedural rather than analytical: take aliquots after each important step and compare the quantity of proinsulin-related polypeptide against the amount present before that step [5]. Final purification in that work used a silica-based reversed-phase matrix in place of a polystyrene-divinylbenzene-based matrix [5].

Designing the cleavage out of the process

One line of work has attacked the enzyme steps directly. A 2025 study engineered a novel mini-proinsulin by shortening the C-peptide and incorporating specific residue substitutions, with the stated aim of eliminating the need for enzymatic cleavage [8]. Production was then optimised by combining microscale high-throughput cultivation in a microbioreactor with statistical modelling using response surface methodology [8].

The medium development followed a two-stage statistical design: critical medium components were screened with a Plackett–Burman design, then refined with a central composite design, which identified glycerol as the most influential factor for yield [8]. Of four statistically derived formulations, the one designated Scenario III gave the highest productivity at microscale, 13.00 g/L, and held up on scale-up to a 3-L bioreactor at 11.5 g/L [8]. The authors describe the optimised medium as balancing carbon and nitrogen sources to support cell viability and protein expression [8], and present the workflow as a model for developing expression systems for other complex recombinant proteins in E. coli [8].

The transferable point for a process engineer is the small gap between microscale prediction and bioreactor result reported in that study [8] — screening at microscale is only useful if the ranking it produces survives scale-up.

The yeast route: secretion, signal sequence, and host strain

Pichia pastoris is a popular yeast for secretory protein production and is an established host for secretory insulin precursor production [2]. Secretion is not automatic: proteins with awkward folding properties, or proteins produced at exaggerated rates, can be retained in the endoplasmic reticulum, triggering the unfolded protein response (UPR) and the induction of endoplasmic reticulum associated degradation (ERAD) [2].

How the precursor is directed out of the cell matters. A comparison of clones using a full-length versus a truncated α-factor secretory signal sequence found the truncated-signal clone (CL4) expressed insulin precursor at a significantly higher level than the full-length clone (HF7) in both media tested [6]. The reported margin was 8.97 times higher in buffered methanol complex medium and 1.17 times higher in methanol basal salt medium [6]. The authors conclude that deleting parts of the secretory signal sequence significantly improved expression efficiency [6].

What the fed-batch profile does to the host

Two studies on controlled fed-batch cultures describe the cell physiology behind the numbers. Cells were grown first in batch on a defined medium with a high glycerol concentration; after glycerol depletion, precursor production was initiated by methanol addition under continuous feeding [2]. The most prominent proteome changes after methanol feeding began were in central carbon metabolism, with the methanol dissimilatory pathway enzymes — virtually absent during the glycerol batch — dominating the proteome in the methanol phase [2]. UPR and ERAD related proteins declined, and extracellular accumulation reached about 3 g/L of insulin precursor without a significant burden on the secretory machinery [2].

A follow-up study showed that reading that decline at face value is a trap [3]. High intracellular levels of the UPR marker chaperone Kar2/Bip were reached during exponential growth on glycerol and then fell sharply on entry into stationary phase, with electron microscopy showing autophagic processes including ER phagy at the end of the glycerol batch [3]. Pre-induction Kar2/Bip levels did not affect secretion efficiency in the subsequent methanol phase [3]. Critically, Kar2/Bip appeared in the culture supernatant of producing cultures but not in non-producing controls, and most of it was present as a truncated protein of approximately 35 kDa [3]. The authors conclude that judging whether a product activates the UPR requires analysing the medium for truncated as well as intact marker protein [3].

Impurities: what the analytics have to find

Whichever route is used, the finished molecule is judged on what else is in the vial. In development batches of recombinant insulin lispro produced in E. coli , ion exchange chromatography revealed an unknown derivative in excess of the assumed limit [4]. Its mass was 42 Da higher than the theoretical mass of Lys(B31) insulin lispro, one of the expected process-related intermediates [4]. Endoproteinase Glu-C digestion localised the modified peptide, tandem mass spectrometry placed the shift on the B31 lysine via y-type ions, and two diagnostic ions at m/z 143.1 and 126.1 indicated lysine acetylation [4]. Edman sequencing against laboratory-synthesised phenylthiohydantoin standards confirmed the residue as N-ε-acetyl-L-lysine [4].

A second degradation pathway is oxidative. Recombinant human insulin exposed for three hours at room temperature to a copper(II)/ascorbate system formed DOPA and a quinone species (DOCH), and Michael addition to those tyrosine oxidation products created new covalent cross-links and aggregates [1]. Cross-links were detected between DOCH at positions B16, B26, A14 and A19 and the free amino groups of the N-terminal residues Phe B1 and Gly A1, plus the side chains of Lys B29, His B5 and His B10 [1]. Fragments from peptide bond hydrolysis were also found [1]. Aggregate content was followed by size exclusion chromatography and the species localised by electrospray ionisation mass spectrometry across different pH and temperature conditions [1].

What to take from this when choosing a process

Three questions separate one insulin process from another. Second, how many enzymatic conversions does the construct require, and can design of the precursor remove any of them [8]. Third, how is the host actually behaving under the feed profile, which the Kar2/Bip work shows cannot be read from intracellular marker levels alone [3].

References

  1. Torosantucci R, Mozziconacci O, Sharov V, Schöneich C, Jiskoot W. Chemical modifications in aggregates of recombinant human insulin induced by metal-catalyzed oxidation: covalent cross-linking via michael addition to tyrosine oxidation products — Pharmaceutical research, 2012-05-10.
  2. Vanz AL, Nimtz M, Rinas U. Decrease of UPR- and ERAD-related proteins in Pichia pastoris during methanol-induced secretory insulin precursor production in controlled fed-batch cultures — Microbial cell factories, 2014-02-13.
  3. Roth G, Vanz AL, Lünsdorf H, Nimtz M, Rinas U. Fate of the UPR marker protein Kar2/Bip and autophagic processes in fed-batch cultures of secretory insulin precursor producing Pichia pastoris — Microbial cell factories, 2018-08-09.
  4. Szewczak J, Bierczyńska-Krzysik A, Piejko M, Mak P, Stadnik D. Isolation and Characterization of Acetylated Derivative of Recombinant Insulin Lispro Produced in Escherichia coli — Pharmaceutical research, 2015-02-07.
  5. Mackin RB. Alternative preparation of inclusion bodies excludes interfering non-protein contaminants and improves the yield of recombinant proinsulin — MethodsX, 2014-08-12.
  6. Utami N, Nurdiani D, Hariyatun H, Putro EW, Patria FP, Kusharyoto W. Full-length versus truncated α-factor secretory signal sequences for expression of recombinant human insulin precursor in yeast Pichia pastoris: a comparison — Journal, genetic engineering & biotechnology, 2023-05-22.
  7. Keane J, Tajouri L, Gray B. Recombinant human growth hormone and insulin-like growth factor-1 do not affect mitochondrial derived highly reactive oxygen species production in peripheral blood mononuclear cells under conditions of substrate saturation in-vitro — Nutrition & metabolism, 2016-07-04.
  8. Ayan E, Aytekin AÖ, Kati A, Demirci H. Optimizing recombinant mini proinsulin production via response surface method and microbioreactor screening — PloS one, 2025-09-08.

All articles