Escherichia coli–expressed near full length HIV-1 envelope glycoprotein is a highly sensitive and specific diagnostic antigen

E. coli host strains DH5α and BL21(DE3), used for cloning and expression, respectively, were purchased from Invitrogen Life Technologies (Carlsbad, CA, USA). Plasmid pET-28a(+) was from Novagen (Madison, WI, USA). Plasmids pTrxBAP and pBirA have been described before [20, 21]. Enzymes for routine cloning work were procured from MBI Fermentas (Burlington, Canada). Ni-NTA super flow resin and isopropyl-β-D-thiogalactopyranoside (IPTG) were from Qiagen (Hilden, Germany), and Calbiochem-EMD Biosciences (La Jolla, CA, USA), respectively. The intrinsically fluorescent isothiocyanate activated nonadentate Eu³⁺-chelate [22, 23] and biotin isothiocyanate [24] for antigen labeling were synthesized as before. Nucleic Acid Purification (NAP) Sephadex G-25 columns were procured from GE Healthcare (Uppsala, Sweden). SA-coated normal capacity low-fluorescent microtiter plates were obtained from Kaivogen Oy (Turku, Finland).

A collection of 131 human sera samples was assembled for the study. This consisted of one in-house panel and 4 commercially procured panels. The in-house panel of 59 sera samples (consisting of 22 HIV positive and 37 HIV negative sera), pre-screened using Vidas HIV Duo Quick kit (bioMérieux SA, Marcy I’Etoile, France), was obtained from the Department of Virology, University of Turku. These sera samples were collected as per the guidelines of Ethics Committee of the Turku University Hospital. Well-characterized worldwide HIV performance panel (WWRB 302–01 to WWRB 302–30), HIV-1 seroconversion panel (PRB 931–01 to PRB 931–09), anti-HIV-1 low titer performance panel (PRB 108–1 to PRB 108–15) and viral co-infection panel (PCA 201–01 to PCA 201–25) were purchased from SeraCare Life Sciences (Milford, MA, USA). In total, there were 77 anti-HIV-1 positive, 2 anti-HIV-2 positive and 52 anti-HIV negative samples.

The 760 amino acid (aa) residue long r-Δgp160 antigen was designed by eliminating the signal sequence (aa 1–43) and two internal hydrophobic regions (aa 519–538 and 676–706) of HIV-1 (strain NL4-3). The corresponding synthetic gene, codon-optimized for E. coli expression was obtained from Geneart (Regensburg, Germany), and used to express two variants of the encoded recombinant antigen. This synthetic gene was inserted into (i) pET-28a(+), in-frame with the vector-encoded 6x-His tag and stop codon at its carboxyl end, to express r-Δgp160 antigen, and (ii) pTrxBAP in frame with the vector-encoded Trx (thioredoxin)-6x-His tag-BAP sequences at the amino terminus, to express the r-Trx-BAP-Δgp160 antigen. The resultant plasmids were introduced separately into E. coli BL21(DE3), and induced to express with 1 mM IPTG. In order to achieve biotinylation of the r-Trx-BAP-Δgp160 antigen in vivo, E. coli cells expressing this antigen were co-transformed with IPTG-inducible biotin ligase expressing plasmid pBirA and cultured in medium supplemented with biotin (10 μg/ml).

As both recombinant protein variants were associated with the insoluble fraction, they were purified from induced cells under denaturing conditions [20]. Briefly, induced cell pellets were re-suspended in lysis buffer (6 M guanidine HCl/20 mM Tris–HCl/300 mM NaCl/10% glycerol/0.1% sodium deoxycholate/1% Tween-20/10 mM β-mercaptoethanol/20 mM imidazole, pH 8) and lysed by sonication at 4°C (Sonics Vibracell sonicator). The resulting lysate was clarified and chromatographed on a 5 ml Ni-NTA super flow resin column. After washing the column (wash buffer was similar to lysis buffer lacking sodium deoxycholate with the exception that 6 M Guanidine-HCl was replaced by 8 M urea), elution was performed using a linear imidazole gradient (20–500 mM) in wash buffer. All the eluted fractions were analyzed by SDS–PAGE, the peak fractions were pooled together, filtered through 0.22 μm membrane and stored in 1 ml aliquots at a concentration of about 1 mg/ml at −20°C until further use. One aliquot was thawed at a time, heat-treated at 80°C for 90 seconds, snap-cooled in ice bath and the buffer was changed to 50 mM sodium carbonate buffer, pH 9.8 using NAP-5 or NAP-10 column, for subsequent labeling reaction.

The r-Δgp160 antigen was labeled with a 30-fold molar excess of Eu³⁺-chelate at 4°C for 16–20 hours in 50 mM sodium carbonate buffer, pH 9.8, essentially as described before [20]. The r-Trx-BAP-Δgp160 (which failed to be efficiently biotinylated in vivo) was in vitro (chemically) biotinylated with a 40-fold molar excess of biotin isothiocyanate. The reaction was performed in 50 mM sodium carbonate buffer, pH 9.8, at room temperature (RT) for 4 hours. Both the unincorporated labels were removed by passing the reaction mixtures sequentially through NAP-5 and NAP-10 columns. For the sake of simplicity, Eu³⁺-chelate labeled r-Δgp160 and chemically biotinylated r-Trx-BAP-Δgp160 antigens are referred to as r-Δgp160-Eu³⁺ and r-Δgp160-Bio, respectively. The labeled antigens were supplemented with 0.1% bovine serum albumin and 0.05% sodium azide, filtered through 0.22 μm membranes and stored at 4°C, until further use.

Either r-Δgp160-Bio antigen (600 ng/100 μl assay buffer/well) or the previously reported [20]in vivo biotinylated r-Bio-HIV-1env antigen (200 ng/100 μl assay buffer/well) was immobilized on SA-coated microtiter wells. The assay buffer was 50 mM sodium carbonate, pH 9.6 (containing 25 mM NaCl/0.1% Tween-20/2.5% BSA/1.25% sucrose/0.06% γ-globulins from bovine blood/0.05% NaN₃). Binding of biotinylated antigen to SA-coated microtiter wells was allowed to proceed for 1 hour at RT with shaking. The wells were then rinsed twice with wash buffer (154 mM NaCl/0.5 M KCl/0.1% Tween-20/50 mM potassium phosphate buffer, pH 7.2). These antigen-coated wells were used in one- and two-step assay formats as described below.

In the one-step assay (performed only with r-Δgp160-Bio antigen-coated well), a mixture of 50 μl serum, diluted (1:10) in assay buffer, plus 50 μl assay buffer containing 100 ng r-Δgp160-Eu³⁺, was added to each well and incubated for 1 hour at RT with shaking. The wells were washed 7 times with wash buffer. Time-resolved fluorescence for Eu³⁺-chelate was measured (λ_ex:340nm; λ_em:615nm) from dry wells using Victor³V 1420 Multilabel counter (modified standard europium protocol; measurement height 5 mm).

In the two-step assay, 100 μl serum (diluted 1:20 in assay buffer) was added into each well and incubated for 1 hour at RT with shaking. Following 4 washes, 100 μl assay buffer containing either 100 ng r-Δgp160-Eu³⁺ (where r-Δgp160-Bio was immobilized as capture antigen) or r-HIV-1env-Eu³⁺ (where r-Bio-HIV-1env was immobilized as capture antigen) was added to each well and incubated for 1 hour at RT with shaking. The wells were washed 7 times with wash buffer and read as above.

Each assay was evaluated with all the sera on the same day and the results obtained from different assays in single analysis were compared. Sera were designated as either positive or negative using cutoff values of 1000 counts per second (cps) and 924 cps, respectively for the one- and two-step r-Δgp160-based assays developed in this study. A cutoff value of 818 cps was used for the two-step r-HIV-1env assay, based on the previous study [20]. These cutoff values were obtained by adding 3 times the standard deviation (SD) to the mean read-out of 37 HIV-negative sera (of the in-house panel) in the one-step r-Δgp160- (mean = 598 cps; SD = 134 cps), two-step r-Δgp160- (mean = 509 cps; SD = 138 cps), and two-step r-HIV-1env- (mean = 577 cps; SD = 80 cps) antigen based assays described above. The use of three times SD to determine the cutoff value was based on a preliminary characterization of the assay with seronegative samples (n=37). Serum samples with ‘signal-to-cutoff’ (S/Co) ratios of <1.0 were designated as negative, while those with S/Co≥1.0 were designated as positive. Statistical significance of differences in sensitivities between assays was assessed using two-tailed Student’s t test. Differences were considered to be of statistical significance when the probability (p) was <0.05.

The two HIV-1 gp160 antigen variants, r-Δgp160 and r-Trx-BAP-Δgp160, created for this study are shown schematically in Figure 1. Both were designed to delete the amino-terminal signal peptide and two internal hydrophobic stretches of the native HIV-1 gp160 molecule without disruption of the reading frame. Additional sequences were added to these through the use of distinct expression vectors. Both were provided with 6x-His tags to facilitate purification. The key difference between the two was that the r-Trx-BAP-Δgp160 variant also contained at its amino terminus a thioredoxin (Trx)-biotin acceptor peptide (BAP) fusion partner. The purpose of the Trx moiety was to promote solubility of the recombinant antigen and thus facilitate its in vivo biotinylation at the target site provided by BAP. The two antigens were expressed by IPTG induction of BL21(DE3) host cells harboring the cognate expression constructs. In the case of r-Trx-BAP-Δgp160, bacterial biotin ligase was co-expressed to mediate biotinylation of its BAP moiety. An SDS-PAGE analysis of aliquots of cells before and after induction of expression of the r-Δgp160 and r-Trx-BAP-Δgp160 constructs is shown in Figure 2. It was difficult to discern the presence of a distinct induced polypeptide of the predicted molecular weights in both cases, though an additional band of ~34 kDa, consistent with the size of biotin ligase, was clearly seen in cells harboring the r-Trx-BAP-Δgp160 construct. Immunoblot analyses of these induced lysates with a monoclonal antibody specific to the 6x-His tag revealed that both r-Δgp160 and r-Trx-BAP-Δgp160 antigens are indeed present in the induced lysates. However, both recombinant antigens were found to be associated with the insoluble fraction (data not shown). Apparently, the Trx fusion is ineffective in promoting solubility of r-Δgp160 antigen. This is consistent with the previous observations that E. coli-expressed Trx fusion antigens are not necessarily soluble [20, 21]. Therefore, we attempted to purify these antigens under denaturing conditions. This resulted in near homogenous preparations of the two antigens, as evidenced by the SDS-PAGE profiles of the pooled peak fractions in Figure 2 (lanes marked ‘P’). The observed mobilities of purified r-Δgp160 and r-Trx-BAP-Δgp160 were consistent with their predicted sizes of ~88 and ~109 kDa, respectively. Our data show that deletion of the hydrophobic regions is compatible with the expression of r-Δgp160 variants in E. coli. Starting from 1L induced E. coli cultures we obtained ~20 and ~15 mg, respectively, of purified r-Δgp160 and r-Trx-BAP-Δgp160. However, as purifications were carried out in the presence of urea, it had to be removed from the purified antigens prior to their labeling for use in the double antigen sandwich assays (below). We did not determine if the purified proteins are correctly folded and retained the integrity of their conformational epitopes. However, the recognition of linear epitopes by their cognate antibodies may be sufficient to develop a sensitive immunoassay. Further, the hydrophobic regions deleted in r-Δgp160 are not exposed in the native structure of the full-length protein and therefore not expected to contribute to immunoreactivity [25]. It is relevant to point out that expression of full-length HIV-1 gp160 in E. coli has not been possible thus far due to instability/toxicity issues related to full-length, in-frame gp160 gene [19]. In mammalian cells, gp160 undergoes proteolysis in the Golgi complex to produce gp120 and gp41. It was shown recently that when gp120 is expressed in the absence of gp41 in mammalian cells it forms aberrant disulfide linked dimers [26]. Further, it has been reported that gp41 which can be expressed in E. coli, is associated with inclusion bodies regardless of whether it has a GST fusion partner [27, 28]. Purification under denaturing conditions was reported to yield ~5 mg [28] to ~12 mg [27] soluble protein/L of bacterial cell culture.

Having obtained the pair of gp160-based recombinant antigens described above, we set up a TRF double antigen sandwich assay [20]. In this instance, r-Trx-BAP-Δgp160 was immobilized onto SA-coated plates and used to capture anti-HIV antibodies, and r-Δgp160-Eu³⁺ was employed in TRF assay to detect the captured antibodies. We have found in earlier work that immobilizing the biotinylated capture antigen on SA-coated plate enhances the sensitivity of the assay [29]. Surprisingly, initial assays revealed that this assay was untenable as it required very high amounts of r-Trx-BAP-Δgp160 antigen to obtain acceptable signal to background ratios. Further investigation showed this to be a result of low levels of in vivo biotinylation of this antigen (data not shown). The reason for this is not clear, though presumably it may be a reflection of the large size of the antigen, about twice as large as the in vivo biotinylated HIV-1 and −2 gp160 deletion variants reported earlier [20], and its intrinsic insolubility. It may also have been due to inaccessibility of the BAP moiety placed between the Trx and Δgp160 components of the chimeric protein. However, we think this is unlikely as similar internal positioning of BAP did not compromise biotinylation in our earlier studies [20, 21]. To compensate for the lack of in vivo biotinylation, we resorted to in vitro biotinylation of this antigen. The resultant in vitro biotinylated derivative, r-Δgp160-Bio, was used (in place of the initially intended r-Trx-BAP-Δgp160) in the double antigen sandwich TRF assay outlined above (Figure 1). Using r-Δgp160-Bio antigen immobilized on SA-coated plates, we tested sera for anti-HIV antibodies using either a one-step or two-step assay protocol. In the former, serum and r-Δgp160-Eu³⁺ antigen were added together into the microtiter well containing SA-immobilized r-Δgp160-Bio, whereas in the latter a wash step was introduced after serum addition, and then followed by the addition of the r-Δgp160-Eu³⁺ antigen. The performance of both these assays utilizing the r-Δgp160 antigen pair above, against a large panel of well-characterized sera (n=131) was analyzed as described below.

First, we carried out the one- and two-step assays outlined above using a panel of 59 in-house human sera that had been pre-screened with the commercially available Vidas HIV Duo Quick kit, shown in Figure 3. Based on the results with this kit, 22 sera were positive and 37 negative for anti-HIV antibodies. Remarkably, these results were mirrored exactly by both the one- and two-step r-Δgp160-based TRF assays, with no discrepancy in any of the samples whatsoever.

The results with the in-house sera panel warranted additional evaluation. For this purpose, we tested several commercially procured well-characterized sera panels described earlier [20]. To begin with, the sensitivities of the one- and two-step assays were evaluated using a HIV-1 seroconversion panel consisting of 9 samples (PRB 931). Only one out of five currently used commercial kits is capable of detecting seroconversion on day 28, represented by panel member #06. The data summarized in Table 1 show that both the in-house assays picked up panel member #06 unambiguoulsy, demonstrating that they are as good as the best performing Abbott HIV-1/2 kit in this regard. The capacity to detect early seroconversion may be attributable to the inherent design of the double antigen sandwich assay format which narrows down the seroconversion window due to its ability to pick up both IgM and IgG antibodies [20, 30]. The two-step assay displayed relatively higher S/Co ratios. Next, we analyzed a worldwide panel (WWRB 302) consisting of 21 HIV-1-positive, 2 HIV-2-positive and 2 HIV-negative sera, from multiple locations in Asia, Africa, Europe, and North and South America. The data are summarized in Table 2. But for panel members #1 and #11, which were missed by the one-step assay, all other HIV-positive samples were picked out by both in-house assays. These two sera, one corresponding to HIV-1 group O (panel member #01) and the other to HIV-2 (panel member #11), were scored as weakly positive by the two-step assay. On the other hand, the HIV-2-positive panel member #25 was detectable with both the assays. Given that our antigen was not designed to pick up HIV-2, these results imply cross-reactivity of r-Δgp160 with anti-HIV-2 antibodies. Importantly, the two sera in this panel (members #10 and #30), which were HIV-negative in five different commercial assays were scored as seronegative by both our in-house assays. We then evaluated the specificity of the in-house one- and two-step assays, using a viral co-infection panel (PCA 201), consisting of 9 (of which 7 were available to us) HIV-negative and 16 HIV-positive sera samples. Many of the panel members were also seropositive for hepatitis B virus (HBV), hepatitis C virus (HCV) and/or Human T cell Leukemia virus (HTLV). Both assays recognized all HIV-1 seropositive samples correctly, irrespective of the presence of antibodies to other viruses, as shown in Table 3. It is noteworthy that of the 7 HIV-negative sera tested, 6 had evidence of other infections such as HCV, HBV or HTLV. Yet, none of these were picked up by either in-house assay, demonstrating the high degree of specificity of these assays.

The data thus far reveal that both the one- and two-step r-Δgp160-based double antigen sandwich TRF assays performed reliably when tested on three different commercially procured sera panels. Recently, we had evaluated these same sera panels in a two-step TRF assay based on a mixture of recombinant HIV-1 and HIV-2 antigens [20]. As in the present study, the earlier reported assay was also in a double antigen sandwich format. At this point, we were interested to compare the sensitivity of the earlier two-step assay with the current one- and two-step assays. To this end, we analyzed an anti-HIV-1 low titer performance panel of 15 members (PRB 108), using all three assays. The earlier assay was done as before in a two-step format, but using a single antigen, r-HIV-1env (biotinylated antigen to capture and Eu³⁺-chelate labeled antigen to reveal), instead of the two antigen mixture. The results of this comparative analysis are summarized in Table 4. Panel member #2 was unequivocally seronegative by the three commercial EIAs: Abbott HIV-1/2, Genetic Systems HIV-1/2 and Organon Teknika HIV-1 tests. None of the three in-house TRF assays showed reactivity with this sample. Panel member #3, a borderline sample which could not be picked up by the r-HIV-1env-based assay, was identified unequivocally by both the one- and two-step r-Δgp160-based assays. It is noteworthy that this panel member is positive for HIV-1 RNA. Another borderline sample, represented by panel member #12, was undetectable by r-HIV-1env-based assay and one-step r-Δgp160-based assay, but identified as weakly positive by the two-step r-Δgp160-based assay. Four samples (panel member #9, 10, 13 and 14) that were HIV-1 RNA positive, but scored as negative by the Organon Teknika HIV-1 EIA kit, were identified as seropositive for HIV-1 by all three in-house TRF assays. Based on the results in Table 4, the order of performance of the three assays was: r-Δgp160-based two-step assay>r-Δgp160-based one-step assay>r-HIV-1env-based two-step assay.

Collectively the data presented so far indicate that the two-step r-Δgp160-based assay performs better as an HIV-1 diagnostic test. To delineate this aspect more clearly, we compared the S/Co values obtained for the entire collection of sera analyzed in this study (n=131), using the one- and two-step r-Δgp160-based assays. This is presented in Figure 4. The mean S/Co value for the HIV positive serum samples (n = 79) using the two-step r-Δgp160-based assay was 83.1 (95% CI: 69–97.2). The corresponding value for one-step r-Δgp160-based assay was 9.7 (95% CI: 7.1-12.3). Between the two assays, the mean S/Co value was significantly higher in the two-step assay than in one-step assay (p<0.001). We interpret this as a manifestation of the high-dose Hook effect [31], wherein sera with high titers of anti-HIV antibody, in the context of the one-step assay, sequester the tracer antigen, r-Δgp160-Eu³⁺, thereby precluding effective double antigen sandwich formation, leading to lowered S/Co values. This anomaly is eliminated in the two-step assay by introducing an intermediate washing step which removes excess antibody after the initial capture step, before addition of the tracer antigen. However, one-step assay saves time consumed in an additional incubation step.