|Home | About | Journals | Submit | Contact Us | Français|
Human immunodeficiency virus type 1 (HIV-1) evolution and changing strain distribution present a challenge to nucleic acid-based assays. Reliable patient monitoring of viral loads requires the detection and accurate quantification of genetically diverse HIV-1. A panel of 97 HIV-1-seropositive plasma samples collected from Cameroon, Brazil, and South Africa was used to compare the performance of four commercially available HIV RNA quantitative tests: Abbott LCx HIV RNA Quantitative assay (LCx), Bayer Versant HIV-1 RNA 3.0 (bDNA), Roche AMPLICOR HIV-1 MONITOR v1.5 (Monitor v1.5), and bioMérieux NucliSens HIV-1 QT (NucliSens). The panel included group M, group O, and recombinant viruses based on sequence analysis of gag p24, pol integrase, and env gp41. The LCx HIV assay quantified viral RNA in 97 (100%) of the samples. In comparison, bDNA, Monitor v1.5, and NucliSens quantified viral RNA in 96.9%, 94.8%, and 88.6% of the samples, respectively. The two group O specimens were quantified only by the LCx HIV assay. Analysis of nucleotide mismatches at the primer/probe binding sites for Monitor v1.5, NucliSens, and LCx assays revealed that performance characteristics reflected differences in the level of genetic conservation within the target regions.
Measurement of viral RNA levels in plasma is central to optimal clinical management of patients infected with human immunodeficiency virus type 1 (HIV-1) (28, 39). Current guidelines recommend the use of viral load assays to assist in decisions related to the initiation of therapy and changes in therapeutic regimens (http://aidsinfo.nih.gov) (38). Significant underquantification of virus has the potential to lead to an inappropriate strategy for patient management (20). Genetic heterogeneity of HIV-1 is one important factor with potential to jeopardize reliability of viral load measurements (1, 8, 12, 23, 24, 29, 42, 46).
Genomic characterization of HIV-1 strains has revealed the existence of three phylogenetically distinct groups: M, O, and N. Group M viruses have been further subdivided into nine subtypes (27, 34). Envelope gene sequences can differ by 20 to 30% between subtypes and up to 50% between groups. Recombination, contributing to further diversification, is now recognized as central to HIV-1 evolution. In areas where multiple groups/subtypes cocirculate, increasing numbers of intersubtype and intergroup mosaic viruses are emerging (27, 30, 34, 36). Sixteen circulating recombinant forms (CRFs) of HIV-1 have been identified (Los Alamos HIV Sequence Data Base [http://hiv-web.lanl.gov]). In a recent evaluation of 4,250 samples from 13 geographic regions, the overall proportion of HIV-1 recombinants exceeded 18% (34).
The geographic distribution of subtypes and CRFs is unequal and ever changing (34, 36). In the United States and Europe, increasing numbers of non-B infections are being identified. A recent study revealed that 2% of HIV-1-positive blood units collected between 1997 and 2000 in the United States were infected with non-B-subtype viruses (10). Several studies have documented the clinical impact on non-B infections in the United States (25, 48, 49). In many European countries, 20 to 30% of new infections are due to non-B-subtype and recombinant strains (4, 5, 17, 18).
All nucleic acid amplification or signal amplification technologies rely on HIV-1 sequence-specific primers and/or probes. Natural polymorphisms occurring in these target regions have the potential to reduce or abolish hybridization, thus compromising reliability of quantification. In most cases, subtype and target sequence information is unknown at the time of viral load assessment, so genetically divergent variants go unrecognized. For this reason, performance of viral load tests should be as group and subtype independent as possible.
The objective of the present study was to evaluate the performance of four commercially available viral load tests on a panel of 97 well-characterized plasma specimens representing HIV-1 group M subtypes A to D, F, G, CRF01_AE, CRF02_AG, intersubtype recombinants, and group O. The evaluation included three tests approved by the Food and Drug Administration: VERSANT HIV-1 RNA 3.0 assay (bDNA), AMPLICOR HIV MONITOR v1.5 (Monitor v1.5), NucliSens HIV-1 QT (NucliSens), and the LCx HIV RNA Quantitative assay (LCx) (not available in the United States). In addition, the degree of nucleotide conservation within primer and probe binding sites was analyzed for Monitor v1.5, NucliSens, and LCx.
HIV-1-infected plasma samples were collected from individual asymptomatic blood donors from December 1998 to June 2001 by the following: (i) Leopold Zekeng, Laboratoire de Santé Hygiène Mobile, Yaoundé, Cameroon; (ii) Lazare Kaptué, Université de Yaoundé, Yaoundé, Cameroon; (iii) Lutz Gürtler, Loeffler Institute, University of Greifswald, D-17498 Greifswald, Germany; (iv) Roberto Badaro, Fundação Bahiana de Infectologia, Bahia, Brasil; (v) Carlos Brites, Universidade Federal da Bahia, Hospital Universitário, Bahia, Brasil; and (vi) Western Province Blood Transfusion Service, Cape Town, South Africa. All specimens were unlinked and collected per local regulations in the country of origin at the time of collection. Plasma samples were divided into aliquots and stored at −70°C until testing. A 97-member panel, anticipated to be representative of genetically divergent HIV-1, was assembled at Abbott Laboratories for the present study. Genetic characterization and viral load test evaluations were performed in parallel. Samples were not previously tested in any of the viral load assays, and each laboratory was blinded to the contents of the panel.
Three regions of the HIV-1 genome were sequenced: gag p24, pol integrase (pol IN), and the env gp41 immunodominant region (IDR). Total nucleic acid was extracted from 200 to 400 μl of plasma using the QIAamp blood kit (QIAGEN, Inc., Valencia, CA). Primers, conditions for reverse transcription (RT)-PCR amplification, and methods for population-based (direct) sequence/phylogenetic analysis have been described previously (44).
The VERSANT HIV-1 RNA 3.0 assay (Bayer Diagnostics, Tarrytown, NY) was performed at the Instituto de Salud Carlos III (Madrid, Spain) according to the manufacturer's instructions. The procedure requires 1.0 ml plasma and targets highly conserved regions of the HIV-1 pol gene (9, 14). Based on the exUSA package insert, the upper limit of quantitation (ULQ) for this assay is 500,000 (5.7 log10) copies/ml, and the lower limit of quantitation (LLQ) is 50 (1.7 log10) copies/ml.
The AMPLICOR HIV-1 MONITOR v1.5 test (Roche Molecular Systems, Branchburg, NJ) and ultrasensitive sample preparation procedure were performed by LabCorp (Research Triangle Park, NC) according to the manufacturer's instructions. The procedure requires 0.5 ml plasma and uses RT-PCR to target the gag p24 region of HIV-1 (16, 41). Reported values for the ULQ and LLQ were 75,000 (4.88 log10) copies/ml and 50 (1.7 log10) copies/ml, respectively.
The NucliSens HIV-1 QT test (bioMérieux, Inc., Durham, NC) was performed at Children's Hospital of Philadelphia (Philadelphia, PA) according to the manufacturer's instructions. The 0.2-ml sample preparation protocol was used, and the assay targets gag p24 (13). The ULQ and LLQ are 5,011,872 (6.7 log10) copies/ml and 80 (1.9 log10) copies/ml, respectively. Primer and probe sequences were provided to R.H. by bioMerieux.
The LCx HIV RNA Quantitative assay (Abbott Laboratories, Abbott Park, IL) was performed at Abbott Laboratories according to the manufacturer's specifications. This competitive RT-PCR assay targets the pol IN region of HIV-1 (26). Using the 1.0-ml sample preparation protocol, the ULQ and LLQ are 1,000,000 (6.0 log10) copies/ml and 50 (1.7 log10) copies/ml, respectively.
In the present study, the term underquantified is used to denote a difference of >1 log10 copies/ml between two tests. If a sample is undetected by one assay, the LLQ for this test is subtracted from the value obtained in the other test to determine the difference.
Ninety-seven HIV-1-seropositive plasma samples collected in Cameroon (n = 47), South Africa (n = 29), and Brazil (n = 21) were utilized for this study. Three independent regions of the genome (gag p24, pol IN, and env gp41 IDR) were PCR amplified and directly sequenced to assign group and subtype. Specimens for which the subtype designation differed between the genetic regions analyzed were provisionally categorized as mosaics. It is conceivable that a subset of the mosaic category represents dual infections. As shown in Table Table1,1, the panel was composed of 95 HIV-1 group M strains representing multiple subtypes (10 A strains, 17 B strains, 29 C strains, 2 D strains, 5 F strains, 1 G strain, 1 CRF01_AE strain, 22 CRF02_AG strains, and 8 mosaic strains) and 2 group O strains. The mosaic category, defined by gag p24/pol IN/env gp41 IDR, included 3 A/G recombinants (2 G/G/A recombinants and 1 A/A/G recombinant distinct from CRF-02_AG), 1 A/G/F recombinant, 3 BF recombinants (1 F/F/B recombinant, 1 F/B/B recombinant, and 1 B/B/BF recombinant), and 1 H/H/A recombinant. All 29 subtype C samples were collected in South Africa. The 17 subtype B, 1 subtype F, and the 3 B/F mosaics were from Brazil, and the balance of the samples originated in Cameroon.
The percentages of HIV-1-seropositive plasma samples quantified by each of the four assays are shown in Table Table1.1. LCx quantified HIV-1 viral RNA in all 97 (100%) of the samples. The viral load concentrations ranged from 2.06 to 6 log10 copies/ml. The bDNA, Monitor v1.5, and NucliSens tests quantified viral RNA in 96.9% (94/97), 94.8% (92/97), and 88.6% (86/97) of the specimens, respectively. Three (3.1%) of the panel members (one subtype C sample and the two group O samples) were below the LLQ of the bDNA, Monitor v1.5, and NucliSens assays. The viral load as determined by LCx for the subtype C sample was 339 copies/ml, and the two group O samples were quantified at 115 and 1,202 copies/ml, respectively. Monitor v1.5 also failed to detect two CRF02_AG samples. Viral loads determined for these two samples by LCx, bDNA, and NucliSens ranged from 898 to 1,700 copies/ml for one sample and from 4,031 to 6,200 copies/ml for the other, greater than 1 log10 differences from results with Monitor v1.5. NucliSens failed to detect eight (8.2%) additional group M samples of various subtype compositions: 1 A sample, 1 C sample, 1 G sample, 2 CRF02_AG samples, 1 A/G/F mosaic, and 2 G/G/A mosaics. These specimens were quantified by the other three assays at 172 to 6,036 copies/ml.
Viral load results were plotted to compare performance of the four tests (Fig. (Fig.1).1). Ninety-four of 97 samples (96.9%) were quantified by both bDNA and LCx tests (Fig. (Fig.1A).1A). The observed correlation coefficient for the 92 specimens quantified within the dynamic range of both assays was 0.876, with a slope of 0.811 and an intercept of 0.904. Eighty-three percent of viral load measurements were within 0.5 log10 RNA copies/ml between tests. Values for 14 samples differed by 0.51 to 1 log10. Of these, one was lower in LCx than in bDNA, whereas 13 were quantified higher in LCx. One Cameroonian G/G/A mosaic sample was underquantified by 1.41 log10 in LCx (955 copies/ml) compared to the result with bDNA (24,547 copies/ml). One subtype C sample and two group O samples quantified by LCx at 339, 115, and 1,202 copies/ml, respectively, were lower than the LLQ in bDNA.
Both Monitor v1.5 and LCx quantified 92 of the 97 (94.8%) samples (Fig. (Fig.1B).1B). The correlation for the 87 specimens within the dynamic range of both assays was 0.689 with a slope of 0.677 and intercept of 1.307. Measurements obtained for 73.9% of samples were within 0.5 log10 copies/ml. Viral load determinations differed between assays by 0.51 to 1 log10 copies/ml for 17 panel members; 8 were higher and 9 were lower in LCx than in v1.5. A Cameroonian G/G/A mosaic specimen was underquantified by 1.80 log10 in LCx (955 copies/ml) relative to the Monitor v1.5 result (60,256 copies/ml). Three specimens (two CRF02_AG specimens and one A specimen) were detected but significantly underquantified by 1.12 to 1.71 log10 in Monitor v1.5 (2.01, 2.91, and 3.34 log10 copies/ml)relative to LCx (3.72, 4.24, and 4.46 log10 copies/ml). In addition, five specimens were measured at lower than the LLQ in Monitor v1.5. Of these, two CRF02_AG specimens and one group O specimen were underquantified by 1.38 to 2.07 log10 copies/ml compared to LCx results (2.99, 3.77, and 3.08 log10 copies/ml, respectively).
Eighty-six (88.6%) members of the 97-member panel were quantified by both NucliSens and LCx (Fig. (Fig.1C).1C). The correlation was 0.811 with a slope of 0.628 and intercept of 1.583 on the 85 samples quantified within the dynamic range of both assays. Viral load measurements for 66.7% of the samples were within 0.5 log10. Differences ranging from 0.51 to 1 log10 were observed for 22 samples; 16 were higher and 6 lower in LCx than in NucliSens. Four specimens (two subtype B specimens and two CRF02_AG specimens) detected by both assays were underquantified by 1.06 to 1.37 log10 in NucliSens compared to LCx (3.38, 4.23, 3.67, and 4.80). Eleven samples (one subtype A sample, two C samples, one G sample, two CRF02_AG samples, three mosaics, and two group O samples) had LCx values of 2.06 to 4.18 log10 copies/ml but were below the NucliSens LLQ (1.9 log10). Of these, six were significantly underquantified (1.08 to 2.28 log10 copies/ml) by NucliSens relative to LCx results: one C sample, one G sample, two G/G/A mosaics, one A/G/F mosaic, and one group O sample.
Ninety-two (94.8%) of the 97 samples were quantified by both Monitor v1.5 and bDNA (Fig. (Fig.1D).1D). The observed correlation for the 87 specimens with viral loads within the dynamic range of both assays was 0.745 with a slope of 0.792 and intercept of 0.659. Viral load determinations for 78.3% of the samples were within 0.5 log10 between tests. Thirteen measurements differed by 0.51 to 1 log10 copies/ml between assays; 10 were higher and 3 lower in v1.5 relative to bDNA. Five samples quantified by both tests were discordant by >1 log10; two subtype C samples and one CRF02_AG were 1.06 to 1.28 log10 higher, and 2 CRF02_AG samples were 1.48 and 1.53 log10 lower, in Monitor v1.5 relative to results in bDNA. Two CRF02_AG samples were below the LLQ of Monitor v1.5 and, based on bDNA measurements of 2.98 and 3.61 log10, were underquantified by 1.25 and 1.91 log10, respectively. Three samples were not detected by either test: one C sample and the two group O samples.
Eighty-four (86.6%) members of the 97-member panel were quantified by both Monitor v1.5 and NucliSens (Fig. (Fig.1E).1E). The correlation for 79 samples within the dynamic range of both assays was 0.680 with a slope of 0.840 and an intercept of 0.582. Viral load results for 60.9% of the samples were within 0.5 log10 between tests. For 21 panel members, viral load differences between assays ranged from 0.51 to 1 log10; 14 results were higher and 7 lower in v1.5 than in NucliSens. Fifteen samples were discordant by more than 1 log10. Four samples (two CRF02_AG samples, one subtype B sample, and one H/H/A mosaic sample) were detected by both tests but underquantified in NucliSens by 1.06 to 1.56 log10: values ranged from 3.86 to 4.78 in Monitor v1.5.
Eight samples were quantified by Monitor v1.5 but were below the NucliSens LLQ of 1.9 log10; seven of them (one A sample, one C sample, one G sample, one CRF02_AG sample, and three mosaics [two G/G/A sample and 1 A/G/F sample]) had Monitor v1.5 viral loads of 2.93 to 4.78 log10 (1.05 to 2.88 log10 difference). Two CRF02_AG samples measured at 3.34 and 4.23 log10 in NucliSens were underquantified by 1.3 log10 copies/ml in Monitor v1.5. Two additional CRF02_AG samples with viral loads of 3.23 and 3.79 log10 in NucliSens were not detected by Monitor v1.5 (1.53 and 2.09 log10 difference between tests). Three samples were not detected by either test: one C sample and the two group O samples.
Eighty-six (88.6%) members of the 97-member test panel were quantified by both bDNA and NucliSens (Fig. (Fig.1F).1F). The correlation was 0.790 with a slope of 0.933 and intercept of 0.349 for the 85 samples with viral loads within the dynamic range of both assays. Viral load measurements were within 0.5 log10 between tests for 70.8% of the samples. Differences in values for 19 specimens ranged from 0.51 to 1 log10 copies/ml; 7 values were higher and 12 values lower in bDNA than in NucliSens. Five samples quantified by both assays were discordant by more than 1 log10: one B/B/BF mosaic and one CRF02_AG sample were 1.08 and 1.25 log10 lower in bDNA, whereas one B sample and two CRF02_AG samples were quantified 1.05 to 1.25 log10 higher in bDNA than in NucliSens. Eight samples were quantified by bDNA but were below the LLQ (1.9 log10) of NucliSens; four were discordant by 1.53 to 2.83 log10: 1 G sample, 1 A/G/F mosaic, and 2 G/G/A mosaics. Three samples were not detected by either test: one C sample and the two group O samples.
Genetic characterization of the plasma panel included sequence analysis of pol IN, target region of the LCx assay, and gag p24, target region of the Monitor v1.5 and NucliSens tests. Nucleotide conservation within primer and probe sites was examined for all three assays. Overall, LCx assay primer and probe regions within pol IN had fewer nucleotide mismatches than was observed for the Monitor v1.5 and NucliSens assays. For group M viruses, LCx had a mean number of total (primers and probe) nucleotide mismatches of 1.4 (range, 0 to 5), whereas Monitor v1.5 and NucliSens assays had a mean of 4.1 (range, 0 to 9) and 7.9 (range, 1 to 17) total mismatches, respectively. At the forward primer, reverse primer, and probe sites, the mean numbers of nucleotide mismatches for group M samples were 0.7, 0.3, and 0.4, respectively, for the LCx assay and 1.8, 0.4, and 2.1 for the Monitor v1.5 test. The mean number of nucleotide mismatches for group M samples in the NucliSens assay was 1.9 at the forward primer, 0.8 at the reverse primer, 1.4 at the capture probe, and 3.7 at the wild-type probe site.
Table Table22 shows the proportion of nucleotide mismatches that occurs at primer/probe binding sites for group M subtypes (A to D, F, and G), intersubtype recombinants, and group O specimens included in this study. For the NucliSens assay, sequence diversity at the wild-type probe-binding site contributes significantly to the total number of nucleotide mismatches. Of group M strains, the number of mismatches at the forward primer site was highest for subtypes A, G, CRF01_AE, and CRF02_AG for the NucliSens and Monitor v1.5 tests. In the LCx assay, the greatest number of mismatches was observed for the forward primer relative to group O sequences (mean and range of 4). For the two group O samples, the LCx assay had 5 total mismatches in contrast to a mean of 19.5 mismatches observed with Monitor v1.5 and 29 mismatches with NucliSens primer and probe sites.
Five group M specimens were underquantified (1.12 to 2.07 log10) in Monitor v1.5 relative to results with LCx. In order to evaluate the molecular basis for underquantification, the number and position of nucleotide mismatches were assessed for the Monitor v1.5 primer and probe sites (29, 46). The results of this analysis are presented in Fig. Fig.2.2. Four CRF02_AG strains had six identical mismatches distributed across the forward primer site. In each case, two mismatches were observed in the probe-binding site. One subtype A specimen had two mismatches within the forward primer and five distributed across the probe site. With the exception of one of the CRF02_AG strains, harboring a single mismatch, no polymorphisms were present within the reverse primer-binding site for these specimens (data not shown).
Genetic variability of HIV-1 has important implications for diagnostic testing, patient monitoring, and therapeutic outcome (45). Molecular assays designed to quantify HIV-1 rely on nucleic acid-based amplification and/or detection of specific target sequences. Consequently, mismatches at primer and probe sites have the potential to reduce the efficiency of hybridization, resulting in underquantification or inability to detect viral RNA (7). Continued evolution and ever-changing global distribution of HIV-1 groups/subtypes make it prudent to monitor the impact of genetic diversity on performance of commercially available viral load tests.
In the present study, we characterized a panel of 97 HIV-1-positive plasma specimens collected from blood donors in Brazil, Cameroon, and South Africa. The panel was selected to provide geographic as well as genetic diversity. In order to increase the likelihood of identifying recombinant HIV-1 strains, three independent regions of the genome (gag p24, pol IN, and env gp41 IDR) were sequenced to derive a subtype assignment. Phylogenetic analysis revealed that the panel consisted of group M subtypes A to D, F, and G, intersubtype recombinants, and group O. Of interest, 32% of the panel members were intersubtype recombinants with the majority (71%) classified phylogenetically as CRF02_AG. Identification of mosaic viruses is of particular relevance when comparing performance of nucleic acid-based technologies that target different regions of the viral genome, i.e., gag p24 for Monitor v1.5 and NucliSens tests and pol IN for the LCx HIV assay. For mosaics, such as CRF02_AG, the target regions are derived from different subtypes; gag p24 is subtype A, whereas pol IN is subtype G. This illustrates the importance of utilizing multiple genomic regions to increase confidence of subtype determination and the value of directly examining target regions of the assays being compared.
The genetically diverse panel of specimens was used to evaluate performance of four viral load tests. LCx, bDNA, Monitor v1.5, and NucliSens assays quantified 100%, 96.9%, 94.8%, and 88.6% of the 97 samples, respectively. Overall, the highest correlation was observed between measurements in the bDNA and LCx assays (0.876) with values for 83.2% of the samples within 0.5 log10. The correlation of results between NucliSens and LCx was 0.811 with 66.7% agreement (within 0.5 log10). Monitor v1.5 and LCx had a correlation coefficient of 0.689 with 73.9% agreement. Compared to bDNA, the correlations of Monitor v1.5 and NucliSens were 0.745 and 0.790, respectively, with 78.3% and 70.8% agreement. The lowest correlation between viral load determinations was 0.680 between the Monitor v1.5 and NucliSens tests. Moreover, only 60.9% of samples differed by less than 0.5 log10. Of note, 15.5% of the panel members were either undetected or discordant by more than 1 log10 in NucliSens, compared to 9.3%, 3.1%, and 1.0% for Monitor v1.5, bDNA, and LCx, respectively.
In several previous studies comparing performance of bDNA v3.0, Monitor v1.5, NucliSens, and LCx (not available in United States) with subtype B specimens, values for correlation and agreement were slightly higher (3, 12, 21-23, 31, 47, 50). The somewhat reduced correlations observed in the present study likely reflect the genetically diverse panel. In studies where performance with non-subtype B virus isolates was examined, correlations and agreement between assays were reduced, comparable with our results (6, 8, 12, 42, 43, 47). Overall, viral load values in the Monitor v1.5 and LCx assays tended to run slightly higher (0.2 to 0.3 log10) than those in the bDNA assay. Similar results have been observed in previous studies (6, 8, 31, 47).
Of the assays evaluated, LCx exhibited the most group- and subtype-independent performance. It was the only assay that quantified all members of the panel. This is consistent with several previous reports indicating that LCx performs well on genetically diverse strains of HIV-1 (12, 42, 43, 47). However, genetic polymorphisms can influence LCx performance, as exemplified by one panel member, a G/G/A recombinant that that was underquantified by >1.0 log10 relative to bDNA and Monitor v1.5. Of group M samples, bDNA, Monitor v1.5, and NucliSens failed to detect one subtype C specimen quantified at 339 copies/ml by LCx. Monitor v1.5 also failed to detect two CRF02_AG samples quantified by the other three assays at 2.95 log10 or greater. Moreover, Monitor v1.5 underquantified (>1.0 log10) two additional CRF02_AG specimens and one subtype A sample. These results are consistent with previous studies where Monitor v1.5 failed to detect or significantly underquantified some strains classified within subtypes A, D, and CRF02_AG (2, 35, 47). Thus, although performance of the Monitor v1.5 assay on non-B subtype viruses is markedly improved relative to that of the Monitor v1.0 assay (29, 42, 43, 46), specific strains within several subtypes and CRFs may be underquantified. NucliSens failed to detect eight additional group M samples of a variety of subtypes: one A sample, one C sample, one G sample, two CRF02_AG samples, and three mosaics (one A/G/F sample, two G/G/A samples). In addition, NucliSens detected but underquantified two subtype B and two CRF02_AG specimens by >1 log10 relative to results of other assays and one H/H/A recombinant (H in target region) by 1.25, 0.8, and 0.79 log10 in Monitor v1.5, bDNA, and LCx. Of the assays examined, performance of NucliSens was most affected by genetic heterogeneity. This is consistent with other studies reporting the inability to detect or underquantitation by NucliSens of specific strains from several group M subtypes (A to C, F, and G), recombinants (CRF01_AE, CRF02_AG), and group O (1, 6, 23, 24, 31-33, 40).
The bDNA assay is generally considered to have quite robust performance characteristics with group M viruses (23, 24, 43). However, cases of failed detection or significant levels of underquantification by bDNA have been noted. For example, Clarke et al. reported one sample that was below the 50-copies/ml LLD of bDNA but was quantified at 4.1 log10 copies/ml by Monitor v1.5 (subtype designation was not reported) (8). Others have demonstrated underquantitation of specific CRF02_AG strains and some subtype F strains (1, 23, 15, 43). In the present study, bDNA underquantified one CRF02_AG sample (>1 log10) relative to Monitor v1.5 and NucliSens and two subtype C strains relative to Monitor v1.5. In addition, one subtype F sample was quantified 0.9 log10 lower in the bDNA assay than in the LCx and Monitor v1.5 assays.
Amendola et al. (2) reported that some CRF02_AG strains are underquantified by LCx relative to other assays. Since the test panel included 22 CRF02_AG samples, we had an opportunity to explore this issue. In the present study, LCx quantified all of the CRF02_AG specimens. Comparison of LCx and bDNA values revealed a high correlation (0.945) with good agreement. Values for only four specimens differed by >0.5 log10, and all were higher with LCx. Correlations between LCx viral loads and those determined by Monitor v1.5 and NucliSens were considerably lower, 0.666 and 0.852, respectively. Values for nine specimens differed by >0.5 log10 between LCx and Monitor v1.5, seven were higher in LCx (four by >1 log10). Viral loads of 10 specimens differed by >0.5 log10 between LCx and NucliSens, 8 were higher in LCx (2 by >1 log10). Thus, LCx and bDNA provided the most robust performance on this set of CRF02_AG samples. The basis for the difference in results obtained between studies is unclear but may be attributable to the monophyletic nature of the CRF02_AG infections in the previous study (1, 2).
Molecular characterization of the panel included sequence analysis of the target regions for NucliSens, Monitor v1.5, and LCx. This provided the opportunity to determine the level of conservation within the primer and probe binding sites for these three assays. For this panel of genetically diverse specimens, the LCx assay exhibited the highest degree of nucleotide conservation. Of the 95 group M pol IN sequences examined, 89.5% of samples had 2 or fewer total nucleotide mismatches at the primer/probe binding sites of the LCx assay. In contrast, only 20% of the gag p24 sequences had 2 or fewer mismatches in Monitor v1.5 primer/probe binding sites and 6.3% for NucliSens. At the LCx forward primer site, 100% of group M samples had two or fewer mismatches, compared to 65.2% for both Monitor v1.5 and NucliSens. Reverse primer sites were well conserved for all three assays, with LCx and v1.5 having 98.9% of samples with two or fewer mismatches and 98.5% with NucliSens. At the LCx probe site, 97.9% of samples had 0 or 1 mismatch, compared to only 20% for the Monitor v1.5 probe region. For NucliSens, 93.7% of samples had one or fewer mismatches at the capture probe site, but the wild-type probe site was considerably less conserved, only 12.6% of samples with one or fewer mismatches.
Group O strains exhibited the most dramatic differences between assays in target region conservation. The LCx assay had a total of 5 primer and probe site mismatches, compared to 19.5 and 29 for Monitor v1.5 and NucliSens, respectively (Table (Table2).2). Thus, relative to LCx, Monitor v1.5 and NucliSens had four to six times as many total primer/probe mismatches for group O specimens, consistent with their inability to quantify group O infections. Unfortunately, a similar analysis could not be performed for bDNA, since probe sequences were not available.
Performance of the LCx, Monitor v1.5, and NucliSens assays reflected the observed differences in the level of nucleotide conservation within the primer and probe target regions. While overall sequence conservation provides some insight into the molecular basis for differences in reliability of quantification, many other factors, such as position, nature and number of mismatches, primer length, and stringency (buffer, temperature, and time) contribute to mismatch tolerance. Few studies have directly investigated the impact of nucleotide mismatches on efficiency of primer/probe binding in the context of viral load assays (7). In an effort to define the molecular basis for underquantification (or lack of detection) in the present study, specific mismatches within target regions were investigated. In many cases, assay performance could be rationalized based on the number and positions of primer/probe mismatches.
For the NucliSens assay, high numbers of mismatches in the wild-type probe-binding site were often identified in missed or underquantified specimens. For example, of the subtype G and G/G/A recombinants (target region subtype G) NucliSens failed to detect, all had seven or eight nucleotide mismatches distributed across the probe region. An undetected A/G/F strain had six mismatches across the wild-type probe. Two underquantified CRF02_AG samples had 12 and 13 total nucleotide mismatches. One CRF02_AG sample had seven mismatches relative to the wild-type probe and five in the forward primer, including one at −2 from the 3′ end. The other had four wild-type probe mismatches and four in the forward primer, with one at −2 from the 3′ end. The relative lack of conservation in the wild-type probe site is a liability for genetically divergent specimens. In a few cases, the molecular basis for failed quantitation was less clear. One subtype B sample had only four total mismatches, none in what might be anticipated as critical positions. A second subtype B specimen (nine total mismatches), originally underquantified by >1 log10, upon retesting yielded a viral load of 4.28 log10, comparable to the other three assays. In several cases, relatively high numbers of mutations located within a primer site were fairly well tolerated by NucliSens. For example, four CRF02_AG specimens were correctly quantified in the NucliSens assay even though six mismatches were distributed across the forward primer. This level of mismatch tolerance in the forward primer presumably reflects the low temperatures associated with the amplification phase of this assay.
One specimen, a unique G/G/A recombinant, was underquantified by LCx relative to results with bDNA and Monitor v1.5 (1.4 and 1.8 log10, respectively). Examination of mismatches revealed a single mismatch in the forward primer (−9 from the 3′ end) and two mismatches distributed evenly across the probe sequence. Analysis using a set of synthetic transcripts containing these mismatches, isolated to specific primer and/or probe regions, revealed that the probe mismatches are responsible for underquantification of this strain (P. Swanson and J. Hackett, Jr., unpublished observation). It should be noted, however, that the presence of two mismatches in the LCx probe region is not necessarily sufficient to result in underquantification. One Brazilian subtype B specimen in the panel also contained two mismatches in the LCx probe site (near the 3′ end) yet was quantified equally by all four assays. Thus, the position of the mismatches within the target region must be considered, as well as the total number and nature of the mismatches. Only 2 members of the 97-member panel contained 2 mismatches relative to the LCx probe site, consistent with previous studies demonstrating a high level of nucleotide conservation within this target region (42, 43).
In the present study, the Monitor v1.5 assay failed to detect two CRF02_AG infections (measured at 2.95 and 3.61 log10 or greater in bDNA and LCx assays) and underquantified two additional CRF02_AG strains by more than 1 log10. Sequence analysis of the primer and probe sites (Fig. (Fig.2)2) revealed that all strains contained two identical genetic polymorphisms in the probe site and six in the forward primer-binding site. Mismatches in the probe site were located at or near the probe termini, leaving a large internal hybridization domain, and therefore are deemed unlikely to be the basis for underquantification. In contrast, the six polymorphisms within the forward primer were dispersed across the binding site, making it unlikely that this primer would hybridize efficiently. This is consistent with results from a previous study examining the impact of internal primer mismatches on performance, where it was shown that five or six mismatches significantly reduce amplification efficiency (7). Probe mismatches also have the potential to compromise reliability of quantification in the Monitor v1.5 assay, as evidenced by underquantification of one subtype A panel member. Sequence analysis revealed the presence of five genetic polymorphisms distributed across the probe site (Fig. (Fig.2).2). It is noteworthy that all five of these underquantified panel members have subtype A-derived target sequences. Development of version 1.5 of the Monitor assay was driven to a significant degree by well-recognized deficiencies of the v1.0 assay in quantification of subtype A strains (42, 46). This is a testament to the ongoing challenges of continual evolution and diversification of HIV-1.
An identical constellation of 6 genetic polymorphisms within the Monitor v1.5 forward primer site in 4 of the 22 (18.2%) CRF02_AG strains in the current study (collected in 1998) prompted us to examine whether these polymorphisms were present in Cameroonian specimens collected in 1996. Four of seventy (5.7%) CRF02_AG strains carried this set of polymorphisms (data not shown). Although the sample size is limited, it appears that the proportion of CRF02_AG strains with six genetic polymorphisms relative to the Monitor v1.5 forward primer may be increasing in Cameroon. This warrants additional study given the widening geographical distribution of this recombinant form. It currently represents the most common source of non-B infections in France and has been detected in the United States and Europe (18, 34).
Of the assays evaluated, only LCx was capable of quantifying the two genetically divergent group O specimens present in the evaluation panel. This is consistent with prior demonstrations of successful quantification of group O strains by LCx (11, 12, 37, 42, 43, 47). However, absolute accuracy is difficult to assess, since no independent group O quantification standard is available. The inability of the Monitor v1.5 and NucliSens assays to detect or reliably quantify group O viruses is in line with results reported by several groups (6, 40, 42, 43, 47). High numbers of nucleotide mismatches present within their primer/probe sites (Table (Table2)2) provide a compelling molecular basis for this observation. Although the bDNA assay can, at least in some cases, detect group O viruses, it consistently underquantifies them (42, 43, 47; V. Soriano, unpublished observation). In this study, the bDNA assay failed to detect the two group O strains, demonstrating that the impact of genetic polymorphisms is not restricted to amplification-based technologies.
The high level of genetic diversity of HIV-1 has important implications for clinical management of infected patients. The influence of genetic heterogeneity on performance was evident for each of the commercial viral load tests evaluated in this study. The relative impact on accuracy of quantification varied between assays. The type of technology used, design of primers and/or probes (conservation of target sequences), and assay conditions all contribute to performance characteristics with genetically polymorphic strains (targets). The use of tests with group/subtype-independent performance offers significant advantages. It is also critical that clinicians maintain vigilance for cases where discrepancies between viral load measurements and the patient's clinical status are evident. Inaccurate viral load determinations have the potential to significantly impact patient outcome (20). The high rate of global diversification of HIV-1 can be expected to continue. Significant advances in molecular technologies will likely be required to meet this challenge. Early indications suggest that application of traditional probes in real-time assays may not prove to be straightforward (19). Thus, it is important to maintain surveillance of HIV-1 genetic diversity and to continually evaluate its impact on performance of diagnostic and patient monitoring assays.
We thank Catherine Brennan, Pierre Bodelle, and Julie Yamaguchi for assistance in acquisition and molecular characterization of the panel and John Robinson and Sven Thamm for critical review of the manuscript.