Insulin or a Closely Related Molecule Is Native to Escherichia coZi*

Evidence is presented for the presence in E s c ~ e ~ i c ~ i ~ coli of material very similar to insulin with regard to specific reactivity in the insulin radioimmunoassay and in the insulin bioassay. prepared from epididymal fat pads of young Sprague-Dawley rats (7). To show that the bioactive molecules were reactive with anti-insulin antibody, duplicate aliquots were mixed with a 1:lOO dilution of guinea pig anti-porcine insulin serum 619 or normal guinea pig serum before their addition to the bioassay. Normal guinea pig serum at this concentration had no effect on the bioassay.

' The possibility of a late recombinant DNA event from a eukaryote source, introducing the insulin gene into E. coli, has not been excluded. extracting insulin from pancreas (1.3). The reconstituted extract was applied to a column of Sephadex G-50 (fine) and eluted with 0.05 M (NW,),CO:,. Each effluent fraction was lyophilized and reconstituted in 1 ml of distilled water.
Radioimmunoassay-The insulin content of each fraction was dekrmined by radioimmunoassay at 1:lO (final) dilution of the sample. Porcine insulin (purchased from Eli Lilly) and guinea pig antiporcine insulin serum ("first antibody"; Batch 619, purchased from the Department of Pharmacology, Indiana University, Indianapolis) were used in the radioimmunoassay. The insulin radioimmunoassay was performed by standard methods using '"I-labeled porcine insulin as the tracer and rabbit anti-guinea pig IgG as the "second antibody" (4, 5).
Exclusion of ~o~s~e c l~e i~y in the Radzoimmunoussay-To exclude the possibility that the gel-fiitered extracts of E. coli contained substances that interfered in a nonspecific manner in the double antibody radioimmunoa~ay. we performed the following experiments. In the first experiment (I), the supernatants from the immunoassay (after removal of the immune precipit.ates) were reacted with trichloroacetic acid at a final concentration of 5%; 92%~ or more of the radioactivity was precipitated irrespective of whether extract had been present. In the second experiment (II), '"I-insulin was incubated for 3 days under conditions of the immunoassay in the presence and absence of extract and then filtered on Sephadex G-50; the profile of the radioactivity was the same irrespective of whether tracer had been exposed to extract. In Experiment 111, "'I-insulin was incubated with or without extract under conditions of the assay for 3 days. Just before addition of the second antibody, an excess of anti-insulin antibody was added (under conditions such that the total concentration of guinea pig y-globulin was effectively unaltered); 90% or more of the radioactivity was precipitated irrespective of whether extract bad been present. In Experiment IV, extract was incubated with a high concentration of '"I-insulin under conditions of the assay for 3 days. The mixture was then diluted (sufficiently so that its insulin content was negligible, Le. true trace) and used in the radioimmunoassay. '2sII-Insulin so exposed was indistinguishable from ""I-insulin incubated without extract. In Experiment. V, extract was added to the immunoassay for human growth hormone, which utilizes guinea pig antibody to human growth hormone with the same carrier guinea pig y-globulin and rabbit anti-guinea pig globulin used in the insulin assay; little or no effect of extract was observed.
If the extract were acting on the rabbit anti-guinea pig globulin (second antibody) to reduce its ability to bind to guinea pig y-globulin or form precipitates, this effect would have been uncovered in Experiments 111 and V. If the extract were acting on the guinea pig yglobulin to reduce its ability to interact with or be precipitated by second antibody or if the extract were acting on guinea pig y-globulin to destroy the antigen binding sites (i.e. insulin or human growth hormone binding sites), the effects would have been uncovered in Experiments I11 and V. If the extracts grossly altered the '"I-insulin (and thereby prevented it from interacting fully with first antibody), this would have been detected in Experiment 111. Putative low molecular weight binding materials (that bind to '"I-insulin prevent it from binding to first antibody but do not change the migration of 'zOIinsulin on Sephadex G-50), while not detectable in Experiment 11, were excluded by previous gel fiitration and dialysis used to prepare the extracts. If the extra;: contained material that did not change the overall features of the &"I-insulin but simply destroyed its affinity (Le. K = 0) for first antibody, this would have been picked up in Experiment IIX. The only remaining possibility is that the extract caused a reduction in (but not total loss off the affinity of '251-insulin for antibody. This possibility, while not entirely excluded by Experiment 111, is partially excluded by Experiment I\'. In summary, we have excluded with reasonable certainty the whole range of possible "nonspecific" effects in the assay.
~ioassay-Fo~owing gel filtration, the fractions corres~nding to the peak of insulin immunoreactivity were pooled, lyophilized, and reconstituted in distilled water. The reconstituted pool was then tested for bioactivity and reassayed for immunoreactivity. The biological activity was measured as lipogenesis, i.e. incorporation of [3-"H]glucose into toluene-extractable lipids To show that the bioactive molecules were reactive with anti-insulin antibody, duplicate aliquots were mixed with a 1:lOO dilution of guinea pig anti-porcine insulin serum 619 or normal guinea pig serum before their addition to the bioassay. Normal guinea pig serum at this concentration had no effect on the bioassay.

RESULTS
When acid ethanol extracts of E. coli K12 were gel filtered on Sephadex G-50 (fine), a peak of insulin immunoreactivity was found in the region typical for insulin ( Fig. 1, inset). On serial dilution over a 6-fold range, the gel-filtered material was very similar to the pork insulin standard (Fig. 1 A ) . About 40 pg/g, wet weight, were recovered from the pooled peak fractions, measured both by radioimmunoassay and by bioassay. The bioactivity was largely neutralized in the presence of anti-insulin antibody (Fig. l B , Table I). The conditioned medium in which the E. coli had been grown also contained @P "f .
The extract was fitered on a column of Sephadex G-50 (fine; 7.5 X 60 cm; 1-ml fractions) and the immunoreactive insulin measured at a 1: 10 final dilution in the assay. The horizontal line, equivalent to 0.25 ng of insulin, indicates the sensitivity of the assay and the fractions that were tested. The arrow at the left indicates the void volume (Iz5Ithyroglobulin) and the arrow at the right marks the salt peak (Iz5I). Peak fractions from the gel fiitration column were pooled, lyophilized, and reconstituted in distilled water, and then tested over a &fold dilution in the immunoassay (X). B, the reconstituted peak fractions were tested for biological activity by measuring the incorporation of  "Column fractions" is the sum of immunoreactive insulin content in peak effluent fractions. Pooled peak fractions were lyophilized, reconstituted, and reassayed in the immunoassay (RIA) and in the bioassay with and without anti-insulin antibody (as described under "Methods" and in Fig. 1). It should be noted that the medium in which the bacteria were grown was composed only of citric acid, salts, and glucose with no serum or other macromolecules; when ( Tetrahymena) medium was carried through the entire procedure and extracted, no insulin was detected ( 1).2 Evidence for Specificity of the Bioassay (Fig. 2)"Insulin and materials that are structurally similar to insulin exercise their bioactivity by binding to specific binding sites of the insulin receptor on the cell surface; the combination of insulin with the receptor activates the bioactivity of the cell. Other materials that have insulin-like bioactivity but are structurally unrelated to insulin often activate bioactivity at steps that bypass the insulin binding site of the receptor (8). As shown in Fig. 2, the insulin-like bioactivities from Tetrahymena and E. coli, as well as that of authentic pork insulin, guinea pig insulin, and an insulin-like growth factor (multiplication stimulating activity) were blocked by antibody directed against the receptor for insulin. This is in contrast to spermine and vitamin Kg which activate the insulin pathway independent of the receptor (8) and had insulin-like bioactivity that was unaffected by anti-receptor antibody. These data indicate that the material in E. coli and Tetrahymena produced their bioactivity by acting through the insulin binding site of the receptor. Note also in these experiments the specificity of the anti-insulin sera; antibodies directed against pork insulin neutralized the effect of pork insulin and of E. coli as well as Tetrahymena extracts but had no effect on guinea pig insulin or the insulin-like growth factors which react very weakly or not at all with this antibody. At the same time, antibody directed against guinea pig insulin neutralized guinea pig insulin but not any of the other insulin-like materials, including that of E. coli and Tetrahymena. These experiments indicate that all of the bioactivity in the E. coli and Tetrahymena extracts was exercised through the insulin receptor In addition to the typical insulin peak, there is a distinct peak of immunoreactive material (which constitutes about 30% of the total immunoreactivity) that elutes slightly later than insulin. The nature of this material is unknown and we have not characterized it further. We have observed it previously in variable proportions in extracts of other unicellular organisms as well as extracts of extrapancreatic tissues of mammals. To our knowledge, this position does not correspond to a known insulin derivative that reacts in this radioimmunoassay, but could represent a modification of insulin, produced by the cell or during the extraction procedure. insulin resistance) blocked the bioactivity of the two extracts (95% neutralization), the two insulins (90-95%), and multiplication stimulating activity (75-100%) but had no effect on the bioactivity of vitamin Ks and spermine, which are thought to stimulate the rat adipocytes at steps beyond the insulin binding site of the receptor. Anti-porcine insulin antibody blocked the effect of porcine insulin (100% neutralization) and of the two extracts (95%) but not that of the other materials (which are known to lack reactivity with antipork insulin antibodies). Finally, as expected, rabbit anti-guinea pig insulin antibody neutralized only guinea pig insulin (75% neutralization). The data for the E. coli extract are the same as those in Fig. 1 and for the Tetrahymena extract as in Ref. 1. and that most of the bioactivity was exercised by a molecular species recognized by anti-porcine insulin antibodies but not by antibodies against guinea pig insulin. The concentration of insulin in E. coli (40 pg/g, wet weight) assumes that E. coli insulin is as reactive as pork insulin in our assay and that we recover 100% of the insulin present in the cell. While insulin at this level is equivalent to 0.1 (or more) of the concentration of insulin in the plasma of fasting adults, it represents only 1 insulin molecule/100 E. coli cells because these bacteria are so small. However, if we assume that E. coli insulin is only as reactive in our assays as is the insulin of a primitive vertebrate (hagfish), the values for E. coli should be multiplied by 40fold. If recovery of insulin from the E. coli is 3%, as we have shown for Tetrahymena, then values for E. coli need to be multiplied again by 30. If concentrations of the hormone (picograms/g) fluctuate (as we have recently found for another peptide in Tetrahymena during growth), then values may need to be multiplied again by 5-10-fold. Overall, the molar concentration of insulin in E. coli may be up to 1,000-10,000fold higher than reported here.

DISCUSSION
Since its introduction in 1960 (4), the radioimmunoassay for insulin has been extensively characterized and its specificity shown to be restricted to insulin. In order to react in this radioimmunoassay, a substance must contain the A and B chains of the insulin molecule joined by disulfide bridges (9, 10). However, amino acid homology is also very important. The insulin-like growth factors (I and 11) and guinea pig insulin have the three characteristic disulfide bridges, have A and B chains that have 50-65% amino acid homology with those of pork insulin, and produce insulin-like bioeffects through the insulin receptor, yet are virtually nonreactive in this radioimmunoassay (11, 12). Likewise, fish insulin with 70% homology to pork insulin reacts with reduced affinity in this radioimmunoassay (9). Since the material in our E. coli extracts reacts specifically in the radioimmunoassay (experiments to rule out nonspecific effects are described under "Methods"), we conclude that the material probably has amino acid homology similar to mammalian insulins, at least in the immunoreactive region of the molecule.
Since the immunoreactivity in our extracts was recovered on gel filtration in the same place as insulin, we suggest that the material in our extracts of E. coli has a size-shape similarity to insulin and is unlikely to be proinsulin. The reactivity of this material in the bioassay and neutralization of this reactivity by the specific receptor blocking antibody suggests that in addition to having that region of the molecule which reacts in the radioimmunoassay, it has the additional regions which are needed for bioactivity, i.e. the region necessary for binding to the insulin receptor (i.e. affinity), and the region necessary for activation of the metabolic pathway (i.e. intrinsic activity). That the bioactivity was neutralized by antiinsulin antibody indicates to us that the bioactive and immunoreactive regions are present on the same molecule. In addition, the neutralization excluded the possibility that the effect was that of an insulin-like growth factor, somatomedin, or other material that can produce insulin-like bioeffects but which lacks the sites that interact with anti-insulin antibody. That the bioactivity/immunoactivity ratio of the material in E. coli is close to unity suggests that it is functionally (and structurally) very closely related to the common mammalian insulins such as pork, beef, human, rat, and mouse. Note that chicken insulin, turkey insulin, and fish insulins have bio/ immunoactivity substantially greater than one, whereas proinsulin and other insulin precursors typically have bio/immunoactivity substantially less than one, and the insulin-like growth factors and other somatomedins (as well as guinea pig insulin) all have no reactivity in the immunoassay, despite their insulin bioactivity which is exercised through the insulin receptor.
The most primitive vertebrate insulin (hagfish insulin) that has been characterized has a bio/immunoactivity ratio of close to unity; both its bioactivity and its reactivity in the immunoassay for pork insulin are markedly reduced (13). Thus, we cannot distinguish whether the insulin in our extracts of unicellular prokaryotes and eukaryotes is closer to hagfish insulin or to a mammalian insulin, because both have bio: immunoactivity ratios of unity. This possibility of poor crossreactivity may in addition explain our seemingly low levels of insulin in unicellular organisms. In addition, since the bacteria in our culture are not homogeneous, we are unable to ascertain whether each individual cell is making insulin.
In addition to demonstrating insulin in unicellular organisms, we have also shown the presence of similar amounts of insulin in the conditioned medium in which the cells were grown. This suggests that the cells are capable of both synthesis and release of insulin, although we have not excluded the possibility that the insulin is released by cell lysis rather than by a biologically meaningful secretory process.
The present finding of material that is very similar to insulin in prokaryotes extends the finding of human chorionic gonadotropic-like material in other bacteria (14), as well as our findings of insulin, adrenocorticotropic hormone, and somato-Insulin or a Closely Related Molecule Native to E. coli statin-like material in T~t r a h~r n e~~. '~ This c o n f i i s our earlier suggestion that. many extrapancreatic (both neural and non-neural) tissues of mammals are capable of making insulin (15).
Although we have not demonstrated a function for these peptides as messenger signals in the unicellular organisms, there are a number of reasons that strongly suggest that these peptides do in fact function at this level: (a) the very strong conservation of the biologically important regions of the molecule; ( 6 ) the known effects of vertebrate hormones (e.g. opiates, catecholamines) (16,17) in protozoa, that is inhibited by specific receptor blocking agents; the in uiuo and in vitro effects of mammalian insulin on E . coli (18); and ( e ) the existence of humoral signalling mechanisms for cell-cell communication by slime mold, yeast, and Myxobacteria (19)(20)(21).