Paraoxonase-1 (PON1) is a mammalian serum protein, the activity of which is related to cardiovascular health and the toxicology of organophosphorus (OP) compounds [1
]. PON1 is thought to be synthesized mostly in the liver, and it is associated with high-density lipoproteins (HDLs) in serum [4
]. The exact function of PON1 is not known, but it is an efficient hydrolase of lactones and esters and an inefficient hydrolase of OP compounds, including pesticide metabolites such as paraoxon (from parathion) and chlorpyrifos oxon, and nerve agents such as sarin, tabun, and VX [1
]. Increased PON1 activity appears to be related to lower levels of oxidation of low-density lipoprotein (LDL) particles, and its hydrolytic activity has been suggested to be directed at oxidized fatty acids and homocysteine thiolactone [6
]. Its increased activity has been shown to be related to decreased atherosclerosis, and it has been implicated in mechanisms of cholesterol efflux [9
]. PON1 also efficiently hydrolyzes bacterial lactones involved in quorum sensing, and it may contribute to innate immunity through this activity [11
]. Although the hydrolysis of OP compounds is almost certainly a promiscuous activity of the enzyme, it contributes to the susceptibility to OP intoxication [12
], and PON1 has been suggested as a lead molecule for a prophylactic or therapeutic bioscavenger of OP toxins [13
]. Human PON1, particularly the R192 alloform, is already sufficiently active to protect against chlorpyrifos oxon and diazoxon exposures without engineering. The turnover of many other OPs by natural PON1 is not sufficient to afford significant protection, but a mammalian chimeric form of PON1 has recently been engineered for significant activity against some G-agents [15
As a result of the physiological and toxicological correlations with increased PON1 activity, there is great motivation to develop PON1 as a therapeutic agent. There are significant difficulties with this: PON1 has only moderate solubility; it has three Cys residues including two forming a disulfide bond, and it is glycosylated [16
]. Human PON1 (huPON1) is very difficult to produce in soluble, folded form in E. coli
. Large-scale fermentation has been used to produce soluble huPON1 successfully in E. coli
, but in poor yields for pharmaceutical production [14
]. This motivated Aharoni and colleagues to generate a chimeric mammalian PON1 by DNA shuffling of mouse, rat, rabbit, and human PON1 isoforms, resulting in a variant called G2E6 that could be expressed when fused to the C-terminus of thioredoxin in good yields in the soluble fraction of E. coli
]. The crystal structure of G2E6 was solved, revealing it to be a six-bladed β
-propeller protein bound to two Ca2+
ions, one of which appears to play a more structural role and one of which is located in what is presumed to be the active-site pocket [16
]. A further generation of DNA shuffling and selection yielded G3C9 PON1, which can be expressed in significant amounts in the soluble fraction of E. coli
without a fusion partner (although it bears a C-terminal hexahistidine tag). Both of the proteins have greatest sequence similarity to the rabbit isoform of PON1, and they differ by 58-59 (G2E6) and 50-51 (G3C9) amino acids from huPON1, depending on the polymorph, mostly on the surface and essentially not at all in the putative active site. (Human PON1 is either Leu or Met at 55, while rabbit is only known to be Leu at this position).
Alignment of PON1 variants. All differences are noted with respect to the human PON1 sequences (Q192/M55 polymorph); the 4E9 sequence and the similar rabbit PON1 sequence are shown in full for reference. Differences between G3C9 and 4E9 are noted in (more ...)
The poor solubility of human PON1 is presumably in part a consequence of its ability to associate with HDL. The nature of this interaction is not clearly defined. PON1 has a signal sequence directing it for cellular export, and it is mutated at the cleavage site for the signal protease, resulting of retention of the hydrophobic signal peptide [18
]. That peptide is disordered in the structure of G2E6 [16
]. There is a large hydrophobic patch on the surface of PON1 that is near the N-terminus, suggesting that this is the HDL interaction surface. The interaction of PON1 with HDL stimulates its activity towards lactones, and removal of the signal peptide (residues 1–20) has been shown to abrogate that stimulation, suggesting that it is critical for proper embedding in the apoA-I HDL particle [20
]. It is not surprising that most of the differences between the very insoluble human PON1 and the more soluble G2E6 and G3C9 are on the surface of the protein, as this is where changes would be most expected to affect the solubility of the folded protein.
Despite this intuitive expectation, surprisingly little is known about how mutations affect the solubility of proteins or how to engineer proteins for greater solubility. Several studies have reengineered membrane proteins to render them soluble. Li and coworkers reengineered phospholamban (PLB), a protein that forms a stable helical homopentamer within the sarcoplasmic reticulum membrane, into a soluble pentameric helical bundle by replacing its lipid-exposed hydrophobic residues with charged and polar residues [21
]. Based on computational design, Slovic and coworkers rationally engineered a water-soluble analog of PLB by changing membrane-exposed positions to polar or charged amino acids, while the putative core was left unaltered [22
]. These constructs were based on the hypothesis that membrane proteins and water-soluble proteins share a similar core and it should be possible to solubilize membrane proteins by mutating only their lipid-exposed residues. The redesigned PLBs mimic all of the reported properties of PLB including oligomeric state, helical structure, and stabilization upon phosphorylation. Based on the same approach, Slovic and coworkers redesigned a water-soluble variant of a membrane protein, potassium channel KcsA, by mutating the lipid-contacting side chains to more polar groups [23
We were interested in determining how mutations to huPON1 would affect its solubility and soluble expression in E. coli. We hypothesized that three different types of mutations might increase the solubility of human PON1. We speculated that (a) removal of the hydrophobic N-terminal leader sequence and (b) mutations of hydrophobic amino acids in the presumptive HDL binding site to polar residues would increase the solubility. We also speculated that (c) the surface residues which were mutated to be more polar amino acids during the directed evolution of G2E6 PON1 were mostly responsible for the increased solubility. To test these ideas, we constructed three mutants of human PON1 called ΔN-huPON1, ΔHDL-huPON1, and g2e6p-huPON1 (). We also combined some of the mutations to look for additive effects on solubility.
To test the solubility of these proteins, we exploited the screen developed by Waldo and colleagues based on fusion of an analyte protein (“protein of interest” or POI) to the N-terminus of the “folding reporter” variant of green fluorescent protein (frGFP) [24
]. Briefly, if the POI folds and is soluble, then the frGFP also folds and its chromophore develops, resulting in fluorescent cells. If the POI is insoluble, then the fusion is found in the membrane-associated fraction and little fluorescence develops. Waldo and colleagues demonstrated that the amount of cellular fluorescence is related to the amount of soluble protein. Consequently, we fused each of the huPON1 variants to the N-terminus of frGFP and determined the fluorescence level of the host bacterial cells. We also examined the activity and stability of the resulting proteins.
We were also interested in the determinants of huPON1 solubility because we wished to use that knowledge to generate variants of engineered PON1 that had significantly greater activity toward OP agents, but with a surface sequence significantly more like native huPON1. Little is known about the immunological effects of the administration of heterologous variants, but the large number of mutations on the surface of G2E6 and G3C9 relative to human PON1 is a cause for concern. In the field of antibody-based therapeutics, human anti-mouse antibody syndrome is a common effect of the administration of mouse-derived antibodies, and so variants of mouse antibodies have been successfully “humanized” by replacing their surface residues with a human sequence while maintaining the binding site residues elicited during affinity maturation [25
]. We speculated that we might be able to “humanize” or at least partially humanize evolved variants of G3C9 PON1 with high OP activity by reverting the surface back to the human sequence, except for solubilizing mutations identified in the first part of this work. We chose to humanize the recently reported 4E9 variant () [15
], which has very high activity against the cyclosarin (GF) analog CMP, using the same strategy used to generate g2e6p-huPON1. Because the substrate specificities of huPON1 and G2E6 PON1 are quite different despite essentially identical active sites, it was not clear if humanization could yield an active enzyme [5
]. Here we show that humanization of 4E9 was successful and suggests a path forward for improved therapeutics based on engineered PON1 variants.