Comparison of DNA Enzyme Immunoassay and Line Probe Assays (Inno-LiPA HCV I and II) for Hepatitis C Virus Genotyping

Abstract

Two methods for genotyping hepatitis C virus (DNA enzyme immunoassay [DEIA] and line probe assay [Inno-LiPA HCV I and II]) were compared on 120 samples and of these 87% were assigned to the same subtype by both assays. There were 15 subtyping discrepancies which involved 5% of type 1 isolates and 90% of type 2 isolates. Amplified products from the core and 5′ untranslated regions (UTR) were sequenced to resolve conflicts. Type 1 discordant samples had a guanosine at position −99 in the 5′ UTR, a characteristic of genotype 1b, and a core region typical of subtype 1a. The eight isolates classified as 2a/2c by LiPA and as subtype 2c by DEIA belonged to type 2.

Hepatitis C virus (HCV) isolates have diverse nucleotide sequences. They have been tentatively classified into six major genotypes and about a dozen subtypes (14). It is generally agreed that the HCV genotype affects infection (10). Genotype 1b is most common in patients who develop active chronic hepatitis C and cirrhosis (13). Patients infected with HCV having this genotype also have higher plasma concentrations of HCV RNA and are less responsive to interferon (7). It is therefore very important to have reliable assays for HCV genotyping. Two methods involving hybridization of PCR products with genotype-specific probes have been recently described: the line probe assay (LiPA) (15) and the DNA enzyme immunoassay (DEIA) (9). We compared these assays, which involve different regions of the HCV genome, by using a panel of sera from French anti-HCV antibody-positive individuals. Amplified products were sequenced, and the sequences were compared with the corresponding sequences from databases to resolve any conflicts.

Sera were selected in 1994 from a consecutive series of 68,492 French volunteers undergoing a routine medical checkup offered by the French national health insurance system. Blood samples were sent to a single laboratory (Institut Régional pour la Santé, Tours, France), where alanine aminotransferase (ALT) activity was determined and compared to a norm (N, equal to the mean value plus 2 standard deviations) calculated for particular sex and age groups after exclusion of values below the 2.5th percentile and above the 97.5th percentile. Samples with ALT activities 20% above N were tested for anti-HCV antibody by a third-generation enzyme immunoassay (Ortho HCV 3.0; Ortho Diagnostic Systems, Raritan, N.J.). It was found that 127 of the 2,327 selected individuals were anti-HCV antibody positive.

HCV RNA was extracted from all sera testing positive in an enzyme-linked immunosorbent assay, as described by François et al. (4). Reverse transcription-nested PCR (RT-nPCR) was carried out for two regions of the HCV genome, the 5′ untranslated region (UTR) and core region with the primers provided by the two genotyping kits (4). After the first PCR, amplified products were analyzed by electrophoresis in a 1.5% agarose gel. nPCR was done when no amplicons were detected. It was found that 121 of the 127 anti-HCV antibody-positive individuals (95.3%) were viremic as demonstrated by RT-nPCR with primers from both the 5′ UTR and the core region. Two genotyping methods were used, LiPA (Inno-LiPA HCV [LiPA-I] kit; Innogenetics, Zwijnaarde, Belgium) and DEIA (GEN-ETI-K DEIA kit; Sorin, Saluggia, Italy) with amplicons from the 5′ UTR and core region, respectively. The LiPA-I kit included 17 probes: generic 1 (2 probes), 1a, 1b, generic 2 (2 probes), 2a (2 probes), 2b (2 probes), 3a (4 probes), and 4/5 (3 probes) (15). The DEIA kit included six probes: 1a, 1b, generic 2, 2a, 2b, and 3a probes. Experimental probes for genotypes 4, 5, and 6 were also evaluated. Sorin recommended that any sample reacting only with probe 2 should be regarded as being of subtype 2c. This information is not included in the kit leaflet.

It was found that 120 of the 121 isolates (99.2%) could be genotyped by both methods. The overall agreement between LiPA-I and DEIA results was satisfactory, with 105 of 120 samples (87.5%) yielding concordant results (Table 1). There were 15 subtyping discrepancies, which involved 5% of type 1 isolates and 90% of type 2 isolates. Four isolates were classified 1b by LiPA-I but 1a by DEIA. One type 1 sample could not be subtyped by LiPA-I, whereas DEIA assigned it to subtype 1b. Nine samples were classified 2a by LiPA-I but 2c by DEIA. LiPA-I gave an ambiguous result (2a/2b) for one isolate, whereas DEIA assigned it to subtype 2b. One sample could not be genotyped, although the patient was viremic and PCR products were obtained for both regions.

TABLE 1.

Comparison of the two genotyping assays

LiPA-I result	No. of samples with indicated DEIA resultb									Totalc
	1a	1b	2a	2b	2c	3a	4	5	1a/3a
1a	24									24/24
1b	4	44								44/48
1nca		1								0/1
2a			1		9					1/10
3						30				30/30
4/5							4	1		5/5
1a/3a									1	1/1
2a/2b				1						0/1
Totalc	24/28	44/45	1/1	0/1	0/9	30/30	4/4	1/1	1/1	105/120

^a1nc, type 1 not assigned to subtype. 

^bNumbers in boldface indicate that the samples were further tested with the LiPA-II kit. One sample was not typeable by either method. 

^cTotal number of concordant results/total number of samples with the indicated result. 

Samples yielding dubious LiPA-I results or results discordant with those of DEIA were further tested with LiPA-II (Inno-LiPA HCV II; Innogenetics). The LiPA-II kit contained 16 probes similar to those of the LiPA-I kit and 5 additional probes, 3 of which were specific for genotypes 4, 5, and 6 (18). LiPA-II reassigned the unspecified type 1 sample to subtype 1b and the 2a/2b sample to subtype 2b. Both results were consistent with those obtained with DEIA. Four of the five isolates assigned to genotype 4/5 by LiPA-I were classified genotype 4 by DEIA; the other was classified genotype 5. The extra probes of the LiPA-II kit gave similar results: four isolates were classified genotype 4c/4d, and one isolate was classified genotype 5a. However, LiPA-II did not resolve any other discordant genotype assignments.

5′ UTR-based LiPA-I and LiPA-II rely on a single nucleotide difference between 1a and 1b subtypes (position −99). This tenuous distinction may lead to the mistyping of 2 to 10% of genotype 1 samples (1, 5, 17). In contrast, the core regions of the 1a and 1b subtypes differ by 17 nucleotides, and DEIA probes cover 4 of these subtype-specific nucleotides. To determine the subtypes of our discordant type 1 samples, we sequenced part of the 5′ UTR (207 nucleotides [−272 to −66]) and the core region (203 nucleotides [159 to 361]) by the dideoxynucleotide chain termination method (12). The sequencing of three of the samples classified as 1a by DEIA but as 1b by LiPA-I and LiPA-II showed that they had a guanosine at position −99 in the 5′ UTR, a characteristic of genotype 1b, but that their core sequences were more similar to the subtype 1a consensus sequence (97%) than to the subtype 1b consensus sequence (90%; P < 0.01).

5′ UTR-based LiPA-I and LiPA-II cannot distinguish the 2a and 2c subtypes because their sequences are identical in this domain (2). In contrast, the core regions of the 2a and 2c subtypes differ by 20 nucleotides in the region covered by the DEIA amplicons. Nine of our isolates were subtyped 2c by DEIA but assigned to 2a by LiPA-I and to 2a/2c (n = 8) or generic type 2 (n = 1) by LiPA-II. The 5′ UTR and core regions of eight samples were sequenced. All eight samples had the same 5′ UTR sequence but had divergent core sequences. A comparison with the consensus 2a and 2c sequences showed that only two samples had sequences more similar to that of 2c than 2a (patients 7 [P < 0.01] and 11 [P < 0.02]). Phylogenetic analyses of these eight samples and 55 published sequences including those of isolates representative of the 2a to 2f subtypes and type 2 sequences of unidentified subtype were carried out by using a neighbor-joining algorithm (11), the maximum-likelihood method (options QFYG; fdnaml program of G. J. Olsen, University of Illinois, Urbana), and the maximum-parsimony method (PAUP version 3.0s for the Macintosh computer; heuristic search). The robustness of each topology was checked by a neighbor-joining method and 500 bootstrap replications. Our isolates clearly belonged to genotype 2 but could not be grouped into the previously defined subtypes (Fig. 1). These data confirm that type 2 is frequently mistyped by rapid-genotyping methods, including LiPA and DEIA (19). Contrary to the recommendations of DEIA’s manufacturer, samples which react only with generic probe 2 should not be considered subtype 2c but unspecified type 2 (6, 20).

FIG. 1.

Unrooted phylogenetic tree of 63 HCV type 2 isolates derived from an analysis of partial core sequences (nucleotides 159 to 361) retrieved from the EBI database or by using the Entrez software from the National Center for Biotechnology Information (about 600 sequences). The consensus sequences of types 2a, 2b, and 2c were obtained from Bukh et al. (2), and sequences 1F.03, 1F.48, and RC.12 were from Cammarota et al. (3). Sequences were aligned manually, and preliminary phylogenetic analyses were used to identify a subset of 120 sequences representative of each monophyletic taxon. More detailed analyses were restricted to type 2 sequences. The topology was obtained by a maximum-likelihood method (double asterisks indicate branches at P < 0.01). Monophyletic taxons also retrieved in the most-parsimonious tree are indicated by a plus sign. Percentages show branches identified according to a bootstrap resampling (500 replicates) by neighbor-joining analysis. Sequences obtained in this study are shaded.

Some reports have assigned new HCV type 2 subtypes after limited phylogenetic analyses performed with only a few sequences (8, 16). Our analysis was based on about one-third of the HCV core region and gave a clear separation of genotypes 2a and 2b. These results are consistent with previous studies covering the entire core region (2). In contrast, subtype 2c seems to be a largely paraphyletic group, making any assignment quite weak. Our isolates that supposedly belonged to subtype 2c intermingled with various other type 2 subtypes such as 2d, 2e, and 2f. Investigations covering other domains of the HCV genome such as E1 and NS5 and more complete phylogenetic analyses involving a large number of type 2 sequences are required for accurate classification of this heterogeneous HCV type.