Line Immunoassay for Confirmation and Discrimination of Human T-Cell Lymphotropic Virus Infections in Inconclusive Western Blot Serum Samples from Brazil

Difficulties in confirming and discriminating human T-cell lymphotropic virus type 1 (HTLV-1) and HTLV-2 infections by serological Western blot (WB) assays (HTLV Blot 2.4; MP Biomedicals) have been reported in Brazil, mainly in HIV/AIDS patients, with a large number of WB-indeterminate and WB-positive but HTLV-untypeable results. Nonetheless, a line immunoassay (LIA) (INNO-LIA HTLV-I/II; Fujirebio) provided enhanced specificity and sensitivity for confirming HTLV-1/2 infections.

each assay, and samples considered reactive or inconclusive in screening were submitted for a confirmatory assay.
Confirmatory assays. A WB assay (HTLV Blot 2.4; MP Biomedicals Asia Pacific Pte. Ltd., Singapore) was used to confirm HTLV-1 and HTLV-2 infection for all the previous studies that generated the samples analyzed here, and these results were interpreted according to the stringent criteria provided by the manufacturer. Briefly, HTLV-1-positive serum samples were defined as having the presence of Gag (p19 with or without p24) and two Env (GD21 and rgp46-I) bands. HTLV-2-positive samples were defined as demonstrating reactivity to Gag (p24 with or without p19) and two Env (GD21 and rgp46-II) bands. Samples that showed the presence of antibodies to both Gag (p19 and p24) and Env (GD21) were defined as being HTLV positive but were considered untypeable. Any other pattern of bands was deemed to be indeterminate.
The present study employed an LIA (INNO-LIA HTLV I/II; Fujirebio, Europe NV, Belgium) in an attempt to confirm and/or discriminate samples with inconclusive results by WB (i.e., WB indeterminate or HTLV positive but untypeable). The strips used in the LIA contain antigens for validation, confirmation, and discrimination. For validation, the line marked by each sample was compared to the control line, and a score ranging from ϩ/Ϫ to ϩ3 was assigned. The confirmatory antigens included Gag p19-I/II, Gag p24-I/II, Env gp46-I/II, and Env gp21-I/II. No bands or the occurrence of a single band (Gag p19-I/II, Gag p24-I/II, or Env gp46-I/II) denoted a negative result. The presence of one band (Env gp21-I/II) or two bands (except Env gp21-I/II) indicated indeterminate results, while two bands (Env gp21-I/II and Gag p19-I/II, Gag p24-I/II, or Env gp46-I/II) indicated HTLV positivity. Three discriminatory bands (Gag p19-I, Env gp46-I, and Env gp46-II) were considered as follows: HTLV-1 positivity was indicated by reactivity to Gag p19-I and/or Env gp46-I, while HTLV-2 positivity was found when samples showed Env gp46-II or a higher intensity of the Env gp46-II band than the Gag p19-I and Env gp46-I bands.
Statistical analyses. Differences in the numbers of males and females in each group were evaluated statistically using the chi-square test. GraphPad Prism software version 5.03 (GraphPad, San Diego, CA, USA) was used for age comparisons between groups using Kruskal-Wallis analysis of variance (ANOVA), complemented with Dunn's multiple-comparison test. Results with a P value of Յ0.05 were considered statistically significant.
Ethical considerations. The present research protocol was approved by the Institutional Review Board of the IAL in São Paulo, Brazil (protocol no. 106D/2012, 62H/2015, and 21I/2016) and by the Institutional Research Board (IRB) of the EBMSP in Salvador, Bahia, Brazil (protocol no. 464.286). All procedures were performed in accordance with the principles established in the Declaration of Helsinki and its subsequent revisions.

RESULTS
The characteristics of the patients (sex and age) and the distribution of WBinconclusive results in each study group are presented in Table 2. More males were found among the HBV/HCV-infected patients in G2, with significant differences in relation to HIV/AIDS patients in G1 and HTLV patients in G3 (P ϭ 0.0048). A comparative analysis of age among the groups showed no significant differences, although the individuals in G2 were older overall. Concerning the distribution of WB-inconclusive samples, the three groups contained more WB-indeterminate than HTLV-untypeable samples: 77.4% in G1, 87.5% in G2, and 64.0% in G3.
The LIA provided confirmation of HTLV-1/2 infection (HTLV-1, HTLV-2, or HTLV) in 66.1% (G1), 83.3% (G2), and 76.0% (G3) of the samples analyzed. Interestingly, most WB-indeterminate results in G1 and G2 were confirmed to be HTLV-2 positive by the LIA, but this was not the case in G3. In G3, only HTLV-1 (40.0%)-and HTLV (36.0%)positive samples were detected among both WB-indeterminate and HTLV-untypeable samples (Fig. 1).  Table 3 shows the WB-indeterminate profiles detected in the present study, the numbers of samples that presented each profile, and the numbers and percentages of samples with confirmed HTLV-1/2 infection. These results indicate that the WB patterns showing Env bands (GD21 and/or rgp46-I or -II) plus p19 or p24 were confirmed as being positive for HTLV-1/2 infection by the LIA. The overall patterns from the WBinconclusive samples (n ϭ 111) and the patterns returned by the LIA are presented in Table 4, revealing that the majority of indeterminate WB profiles that were not confirmed as being positive for HTLV-1/2 infection by the LIA presented only Gag bands in G1, only GD21 in G2, and one of three bands (GD21, rgp46-II, and p24) in G3. It is noteworthy that among 26 HTLV-positive but untypeable samples (14 in G1, 3 in G2, and 9 in G3), after the LIA, 28.6% were confirmed as being HTLV-1 positive, 28.6% were confirmed as being HTLV-2 positive, and 42.8% remained HTLV untypeable in G1. In G2, 66.7% were confirmed as being HTLV-1 positive and 33.3% were confirmed as being HTLV-2 positive, and in G3, 66.7% were confirmed as being HTLV-1 positive and 33.3% remained HTLV untypeable ( Fig. 1 and Table 4).

DISCUSSION
HTLV-1-and HTLV-2-seroindeterminate WB results are prevalent worldwide, with rates fluctuating according to country and study group (geographic areas and populations where the disease is or is not endemic). Several attempts have been made to improve WB sensitivity and specificity, such as adding HTLV-1 and HTLV-2 recombinant envelope proteins and transmembrane protein to the HTLV-1 viral lysate. These include rgp46-I, rgp46-II, and GD21, the latter of which blocks the cross-reactivity of gp21 with Plasmodium falciparum in regions where malaria is endemic (22,23). In spite of these efforts, Blot 2.4 continues to yield high rates of WB-indeterminate and/or untypeable HTLV results (4-8, 10, 11, 13-17, 21). The present study found similar results and disclosed that in populations presenting a high risk of acquiring viral infections (G1 and G2) as well in the general population (G3) of Brazil, a large number of WB-inconclusive results were detected. Several hypotheses were taken into consideration for these WB-inconclusive results, such as low HTLV-1 and HTLV-2 proviral loads, mutations in the provirus (defective particles), low-level production of viral antigens consequently leading to a low level of specific antibody production, the seroconversion period, crossreactivity with other antigens or viruses, coinfection with HIV, and the use of antiretroviral therapy, among others (5, 12-15, 24-27).
Although PCR assays presented lower sensitivity than WB in detecting true HTLV-1/2 infection in HIV-HTLV-coinfected individuals in São Paulo, Brazil, the molecular assays were able to confirm and discriminate between HTLV-1 and HTLV-2 in some WBindeterminate and HTLV-untypeable cases, indicating that both serological and molecular assays are useful for HTLV diagnosis (13,15). Due to the presence of large numbers of WB-indeterminate samples, coupled with the high cost of obtaining WB assays and LIAs in Brazil, we recently proposed an algorithm that employs qPCR to confirm HTLV infection, followed by testing any PCR-negative samples by WB or an LIA. This strategy was shown to reduce costs and improve the diagnostic accuracy of HTLV-1/2 detection (13,15). Nonetheless, due to highly divergent socioeconomic conditions among different regions in Brazil, in laboratories without the means to perform molecular assays, high-performance serological testing presents an acceptable alternative.
Some studies of HTLV diagnosis conducted in blood donors in Latin America (considered an area where HTLV-1/2 is endemic) have reported differing numbers of WBindeterminate samples, which were subsequently confirmed to be positive by PCR (29)(30)(31). Also, in blood donors from another area where the disease is endemic, northeast Iran, WB-indeterminate samples were found to be positive by PCR, and the most prevalent WB bands presented various combinations of rgp46-I, GD21, and gp21 (32).
In corroboration with these findings, the majority of serum samples here that presented WB patterns of GD21 and/or rgp46-I or -II plus p19 or p24 were subsequently confirmed to be positive by the LIA. These types of WB patterns were observed in G1, and PCR assays demonstrated HTLV-1 or HTLV-2 positivity (data not shown) (15). In addition, the majority of serum samples that presented only Gag bands in the WB analysis were negative for HTLV-1/2 infection by the LIA. Of note, one blood sample from G1 that showed a faint GD21 band in the WB analysis tested negative for HTLV-1/2 by both the LIA and PCR. Another serum sample presenting a GD21, rgp46-I, and rgp46-II WB pattern was found to be negative by the LIA; unfortunately, this sample could not be analyzed by PCR because only serum was sent to the laboratory for analysis. However, retesting of this serum sample by WB and the LIA confirmed the discrepant results. It is interesting to note that in serum samples (n ϭ 14) that tested HTLV untypeable by the LIA, PCR confirmed the presence of HTLV-1 in 5 samples and HTLV-2 in another 2 samples from G1 (data not shown), emphasizing the need for employing molecular assays to confirm an HTLV diagnosis in patients with HIV-HTLV coinfection.
In G2, the majority of WB-indeterminate patterns presented either GD21 alone or this protein in association with one Gag or envelope band. The LIA confirmed the presence of HTLV-2 in 11/21 (52.4%) of the WB-indeterminate samples. The high number of HTLV-2-positive samples in G2 leads us to suppose that these patients acquired HBV and HCV, as well HTLV-1/2 and HIV, at the same time, probably by the parenteral route and prior to the time when serological testing for HIV and HBV (1989), and subsequently for HTLV and HCV (1993), became mandatory in blood banks throughout Brazil; in addition, intravenous drug addiction was more frequent in this country, as previously described (18)(19)(20). Corroborating this hypothesis, older age and male sex predominated in G2. Regarding the lack of WB detection of truly HTLV-2infected samples, we have described this difficulty since 2000 (8,13,15) and hypothesized that the rgp46-II (K55) present in the WB strip is not as sensitive for the detection antibodies to the HTLV-2 strains that circulate in Brazil (HTLV-2a subtype, variant 2c) (33). This seems not to be the case for the gp46-II present in the LIA strip.
Concerning the two WB-indeterminate samples in G2 with negative HTLV results by the LIA, both presented reactivity for GD21 in the WB analysis, and one of the samples showed a faint band. Curiously, the LIA demonstrates the best performance in this group of patients, with 20/24 (83.3%) of HTLV-positive samples being detected among the WB-inconclusive samples. Unfortunately, only plasma/serum samples from these patients were available for analysis, which did not allow the use of PCR to perform a comparative analysis of serological and molecular results. Nonetheless, associations between HTLV-1/2 and hepatitis B and C have been reported in several studies conducted in Brazil and elsewhere (18-20, 24, 29, 34).
The WB-inconclusive patterns in G3 were quite different from those in the other groups analyzed. Several of the HTLV-untypeable samples demonstrated the presence of almost all bands corresponding to the HTLV-1 viral lysate, without reactivity to rgp46-I, and six of nine were subsequently confirmed to be HTLV-1 positive by the LIA. Twelve out of the 25 WB-inconclusive samples that could be submitted for PCR (9 WB indeterminate and 3 HTLV indeterminate), 11 were confirmed to be HTLV-1 infected, 6 of which were HTLV untypeable by LIAs (data not shown). In addition, the serum samples that tested negative by LIAs presented three different WB-indeterminate patterns: (i) GD21, (ii) p24 (faint band), and (iii) rgp46-II (faint band). Only one of these samples could be tested by PCR and presented an HTLV-negative result (data not shown). In summary, the samples in G3 were confirmed to be positive for HTLV-1 or HTLV but not HTLV-2 infection. This finding could be partially related to the ethnic origin of the included individuals (African descendants), the lack of HIV infection in this group, and the characteristics of the patients seen at HTLV outpatient clinics in Salvador, BA (35).
Of note, the reasons described above to explain WB-inconclusive results could also be applied to the PCR-negative results in truly HTLV-1/2-infected individuals, including the low proviral loads in HIV/AIDS patients undergoing antiretroviral therapy in G1 and, in some cases, in G2 (13)(14)(15); the observation that HTLV-2 infection is known to show low proviral loads (11,12,25); the presence of defective provirus not detected by the primers employed in the PCR assays (27); and infection with other viruses, such as HTLV-3 or HTLV-4, which can be detected only by using specific primers (36,37).
In fact, we could not rule out HTLV-3 and HTLV-4 infections in Brazil, since populations in central Africa migrated from Africa and Australia to the American continent before Asiatic population migration, and their descendants, such as the Amerindians, could maintain such viruses or spread them to the general population, which could justify the frequent presence of WB-indeterminate results in Amerindians, as previously described (42).
In relation to the LIA, although it presented the best performance in detecting HTLV-1 and HTLV-2, we could not exclude misclassified positive results, as occurred in HTLV-3-and HTLV-4-infected individuals, who could be erroneously diagnosed as being infected with HTLV-2 (36)(37)(38).
Of note, despite the fact that the LIA demonstrated better performance than WB in the serological diagnosis of both HTLV-1 and HTLV-2, additional considerations are warranted for both assays. With respect to WB, we consider the lack of an ability to score the intensity of a positive band to be a problem, since it is not known when a faint band should be considered truly positive. The criteria (band profiles) established by the manufacturer to confirm HTLV-1 and HTLV-2 infections in WB assays are excessively stringent and deserve a review. Taking into account the results obtained here, we suggest that samples presenting only one Gag band (p19 or p24) plus GD21 and rgp46-I or rgp-46-II should be considered HTLV-1 and HTLV-2 positive, respectively, since samples that demonstrated p24, GD21, and rgp46-I bands were confirmed to be HTLV-1 positive by the LIA and PCR. In contrast, we detected true HTLV-2 positivity in samples that showed p19, GD21, and rgp46-II bands. In addition, when Gag bands were undetectable but both envelope bands (GD21 and rgp46-I or rgp46-II) were present, it was impossible to rule out true HTLV-1 or HTLV-2 infection, since seroconversion could be taking place. Indeed, when all bands showed reactivity to HTLV-1 viral lysate antigens, even in the absence of rgp46-I, it was possible to confirm HTLV-1 infection.
In conclusion, the LIA was shown to be the best serological test for confirming HTLV-1 and HTLV-2 infections, regardless of whether individuals were HTLV monoinfected or coinfected. We further highlight the need to review some WB criteria based on our results and those reported previously by others. It remains to be determined whether the superior performance of the LIA was due to the less stringent criteria employed than for WB. Further studies are necessary to confirm these results in a variety of risk populations from Brazil and elsewhere.