High incidence of treatment-induced and vaccine-escape hepatitis B virus mutants among human immunodeficiency virus/hepatitis B–infected patients


  • Potential conflict of interest: Dr. Zoulim consults for and received grants from Gilead, Bristol-Myers Squibb, and Roche. Dr. Lascoux-Combe consults for and received grants from Janssen. She also consults for Gilead and received grants from Roche.

  • This study was supported by grants from the Agence Nationale de Recherche sur le Sida et les Hépatites (ANRS), SIDACTION, and the European Community (HepBvar project contract QLRT2001-00977 and VIRGIL network contract LSHMCT- 2004-503359). Gilead Sciences, Inc., provided an unrestricted grant for the French HIV-HBV cohort and was not involved in any part of the data collection, analysis, or manuscript writing. A. B. was awarded a postdoctoral fellowship from the ANRS.

Address reprint requests to: Karine Lacombe, M.D., Ph.D., Service de maladies infectieuses, Hôpital Saint-Antoine, 184 rue du Faubourg Saint-Antoine, 75012 Paris, France. E-mail: karine.lacombe@sat.aphp.fr; fax: +33 1-49-28-21-49; or Fabien Zoulim, M.D., Ph.D., Centre de recherche en cancérologie de Lyon INSERM U 1052/CNRS UMR 5286, 151 cours Albert Thomas, 69000 Lyon, France. E-mail: fabien.zoulim@inserm.fr; fax: +33 472-68-19-71.


Anti–hepatitis B virus (HBV) nucleos(t)ides analogs (NA) exert selective pressures on polymerase (pol) and surface (S) genes, inducing treatment resistance and increasing the risk of vaccine escape mutants. The rate of emergence for these mutations is largely unknown in patients coinfected with human immunodeficiency virus (HIV) and HBV undergoing dual-active therapy. In a 3-year, repeat-sampling, prospective cohort study, HBV viral genome sequences of 171 HIV-HBV coinfected patients, presenting with HBV viremia for at least one visit, were analyzed every 12 months via DNA chip. Logistic and Cox proportional hazard models were used to determine risk factors specifically for S gene mutations at baseline and during follow-up, respectively. HBV-DNA levels >190 IU/mL substantially decreased from 91.8% at inclusion to 40.3% at month 36 (P < 0.001), while lamivudine (LAM) or emtricitabine (FTC) use remained steady (71.9%) and tenofovir (TDF) use expanded (month 0, 17.5%; month 36, 66.7%; P < 0.001). The largest increase of any mutation class was observed in l-nucleoside–associated pol gene/antiviral-associated S gene mutations (cumulative incidence at the end of follow-up, 17.5%) followed by alkyl phosphonate-associated pol-gene (7.4%), immune-associated S gene (specifically any amino acid change at positions s120/s145, 6.4%), and d-cyclopentane–associated pol-gene mutations (2.4%). Incidence of l-nucleoside–associated pol-gene/antiviral–associated S gene mutations was significantly associated with concomitant LAM therapy (adjusted hazard ratio [HR], 4.61; 95% confidence interval [CI], 1.36-15.56), but inversely associated with TDF use (adjusted HR/month, 0.94; 95% CI,0.89-0.98). Cumulative duration of TDF was significantly associated with a reduction in the occurrence of immune-associated S gene mutations (HR/month, 0.88; 95% CI, 0.79-0.98). No major liver-related complications (e.g., fulminant hepatitis, decompensated liver, and hepatocellular carcinoma) were observed in patients with incident mutations. Conclusion: Vaccine escape mutants selected by NA exposure were frequent and steadily increasing during follow-up. Although the high antiviral potency of TDF can mitigate incident mutations, other antiviral options are limited in this respect. The public health implications of their transmission need to be addressed. (Hepatology 2013;53:912–922)


amino acid




covalently closed circular DNA


confidence interval






hepatitis B surface antigen


hepatitis B virus


hepatocellular carcinoma


human immunodeficiency virus


hazard ratio


interquartile range




nucleos(t)ides analogs


polymerase chain reaction







The broad use of combined antiretroviral therapy for human immunodeficiency virus (HIV)-infected patients has ushered in a new era of decreased acquired immunodeficiency syndrome–related morbidity and mortality. Since then, a new spectrum of chronic diseases, namely hepatitis B virus (HBV) infection, has emerged in this population, causing substantial mortality.[1] Decompensated cirrhosis and hepatocellular carcinoma (HCC) are a growing cause of end-stage liver disease in HIV/HBV-infected patients[2] and are highly associated with long-term HBV replication.[3] Optimal control of HBV replication during HIV infection is therefore needed to prevent these types of complications.

Oral antiviral agents with dual anti-HIV and anti-HBV activity, such as the nucleos(t)ides analogs (NA) lamivudine (LAM), emtricitabine (FTC), and tenofovir (TDF), have been largely used in HIV/HBV-coinfected patients. However, they exert selective pressures resulting in HBV polymerase (pol) gene mutations that confer resistance to other NA.[4] Due to the overlapping nature of genes on the HBV genome, pol gene mutations may also induce mutations on the surface (S) gene. HBV strains with conserved site changes on positions s120, s145, s164, s195 and s196, that is Antiviral Drug-Associated Potential Vaccine Escape Mutants, are associated with decreased antigenicity or immunogenicity of hepatitis B surface antigen (HBsAg)[5, 6] and a reduced binding to anti-HBs antibody,[7] ultimately leading to vaccine failure in chimpanzees.[8]

Historically, NAs with anti-HBV activity have been developed in the auspices of HIV-treatment and most HIV-HBV coinfected patients have been given years of subobtimal anti-HBV therapy.[9] Furthermore, in developing countries with high HBV prevalence, such as those from Sub-Saharan Africa, HBV screening and monitoring are lacking and HIV-HBV coinfected patients often exhibit continuous HBV-replication under suboptimal HBV treatment.[10] Despite the potential clinical and public health implications of HBV-infections with pol and S gene mutants, their rate and determinants have never been properly studied in an optimal longitudinal setting. The major pitfall of previous studies is that genetic variability is mostly determined cross-sectionally, when the patient has a rebound under a specific therapy, while rarely extending follow-up to account for mutation stability or suppression under alternative therapeutic strategies. The goal of this prospective, repeated sample cohort study was to determine the prevalence and incidence of the aforementioned pol and S gene mutations, as well as risk factors related to their emergence, within a well-defined cohort of HIV/HBV-coinfected patients treated with a wide variety of anti-HBV treatments.

Patients and Methods


 Patients were selected from the French HIV-HBV Cohort Study as described.[11] Briefly, this prospective study recruited 308 patients from seven centers located in Paris and Lyon, France, between May 2002 and May 2003. Patients were included if they had HIV-positive serology confirmed by western blotting and HBsAg-positive serology for at least 6 months. All data collection continued until 2010-2011.

For this study, subjects were selected if they had an HBV-DNA viral load >190 IU/mL for at least one time point (n = 215). Among them, subjects were removed because of failure to amplify the HBV genome at all visits (n = 38) or unavailable serum samples (n = 6). Thus, 171 patients were included in the present analysis. All patients provided written informed consent to participate in the cohort and the protocol was approved by the Pitié-Salpêtrière Hospital Ethics Committee (Paris, France) in accordance with the Helsinki Declaration.

HBV-DNA viral load was separately quantified using a commercial polymerase chain reaction (PCR)-based assay (COBAS AmpliPrep/COBAS TaqMan, detection limit: 12 IU/mL; or COBAS Amplicor HBV Monitor, detection limit: 60 IU/mL; Roche Diagnostics, Meylan, France).

Quantification of HBV Mutations

HBV Genome Sequencing by DNA Chip

HBV- DNA was analyzed at enrollment and at months 12, 24, and 36. After extraction of viral DNA (Supporting Information), the whole HBV genome was amplified by a duplex PCR-generating amplicons of 1,509 and 1,721 base pairs. Nonamplifiable samples were reprocessed via nested PCR, allowing amplification of a 741–base pair fragment covering reverse-transcriptase and HBs regions. After labeling, cleavage, and purification, amplicons were subjected to a DNA chip analysis. An Affymetrix DNA chip array (bioMérieux) was used in this study to detect HBV genome mutations and define viral strain genotype. The Affymetrix GS3000 scanner was used to detect fluorescence signals, and results were analyzed with the DNA-Chip Experiment Manager software (bioMérieux).[12]

Classification of HBV Mutation Groups

Mutations were categorized by like characteristics.[4] The first group referred to alkyl-phosphonate–associated mutations on the pol gene, consisting of any conserved site changes at position rt181, specific amino acid (aa) change at position rtA194T, rtN236T, or rtA181T+rtN236T. The second group referred to d-cyclopentane mutations on the pol gene, typically associated with entecavir (ETV) resistance, and was identified by specific aa changes at the following positions: rtT184, rtS202, rtI169T, and/or rtM250I/V with rtL180M+rtM204V/I. The third group referred to l-nucleotide–associated mutations on the pol gene and, due to the overlapping structure of the HBV genome, was also termed antiviral-associated S gene mutations. These mutations were defined as any specific aa change on the pol gene (corresponding S gene mutation) at positions rtM204I (sW196*/S/L), rtM204V (sI195M), rtL180M+rtM204V/I (no additional change from rtL180M mutation), or rtV173L+rtL180M+rtM204V/I (sE164D+sI195M).[7] The final group referred to immune-associated mutations of the S gene, which were defined as any aa change at positions s120 or s145.[13]

Available Data for Specific HBV Mutations and Missing Data

Mutation analysis was available for the first 3 years of follow-up. The first DNA chip used (for inclusion visit) only gave information at positions rt173, rt180, rt204, s120, and s145. For all visits thereafter, additional results were available for aa residues on the rt domain at positions rt169, rt181, rt184, rt194, rt202, rt236, and rt250. Furthermore, 120 samples for which results at position rt236 were indeterminate via DNA chip analysis were processed using population sequencing (Supporting Information). As a result, the total number of patients considered in analyses varied for each endpoint (number of patients per position): rt173 (n = 170); rt180 (n = 170); rt204 (n = 169); rt169, rt181, rt184, rt194, rt202, rt236, and rt250 (n = 119); s120 and s145 (n = 168).

Since HBV sequencing cannot be performed under low HBV replication, a number of time points had missing HBV genetic information (72 samples [10.5%]). In order to more appropriately estimate mutation presence (Supporting Fig. 1), HBV information visit at inclusion was carried over until the last follow-up visit or next available point at which HBV sequencing could be performed. In patients without HBV genetic information at cohort inclusion yet with a rebound in HBV replication during follow-up (n = 27), wild-type HBV was assumed to be the dominant quasispecies prior to rebound.

Total Intracellular HBV-DNA and Covalently Closed Circular DNA Quantification

 DNA was extracted from biopsy specimens using the MasterPure DNA purification kit (Epicentre) according to the manufacturer's instructions. Briefly, 500 μL of lysis buffer was added to liver tissue and then gently ground using a plastic grinder. After an overnight proteinase K digestion at 42°C and RNase digestion (30 minutes at 37°C), 500 μL of precipitation buffer was added to each sample. Samples were well homogenized prior to a 10-minute centrifugation (10,000g). The adequate volume of isopropanol was added to the supernatants and DNA was precipitated. Pellets were rinsed with ethanol 70° and resuspended in 50 μL of Tris–ethylene diamine tetraacetic acid buffer. Intrahepatic covalently closed circular DNA (cccDNA) and total HBV-DNA were quantified via real-time PCR using a LightCycler instrument (Roche Diagnostics, Mannheim, Germany) as described.[14]

Intracellular data from biopsies obtained at inclusion and end of follow-up were only available in 14 patients from the original HIV-HBV cohort (n = 308). Biopsies were not obtained on the entire cohort and were performed under physician discretion using European Association for the Study of the Liver consensus statements for HBV-monoinfected patients (Supporting Information). Of these patients, only five had serum HBV-DNA levels high enough to allow HBV sequencing.

Treatment Data and Adherence

 Data on antiviral treatment were obtained from medical records and confirmed with the treating physician. Treatment adherence was strictly evaluated among patients prescribed TDF while exhibiting virological rebound/nonresponse. For these patients, plasma TDF was quantified using high-performance liquid chromatography with tandem mass spectrometry[15] (Polar-RP Synergi, 2.0 × 50 mm, reversed-phase column; detection limit: 10 ng/mL).

Statistical Analysis

 The numbers of patients with HBV-DNA >190 IU/mL, treated with LAM/FTC, adefovir (ADV), or TDF over time were tested using a nonparametric test for trend. Point prevalence of each mutation class was given for every cohort visit in order to illustrate population trends. We then determined cumulative incidence rates from the remaining patients with continued follow-up and without mutations at inclusion.

A composite endpoint was created for (1) antiviral-associated S gene/l-nucleoside–associated pol gene or (2) immune-associated S gene mutations. These mutations were chosen for their potential public health impact and large frequency at inclusion and during follow-up. Risk factors were examined and separated according to inclusion (logistic regression) or during follow-up (Cox proportional hazards regression). Demographic, treatment, and clinical characteristics related to HIV/HBV infection with P ≤ 0.2 in univariate analysis were retained and used to create a predictive, multivariable model. A backward-stepwise selection process was then performed, removing any nonsignificant covariables (univariate analysis can be found in Supporting Table 1).

All analyses were performed using STATA statistical software (version 11.2), and P < 0.05 was considered significant.


Study Population and HBV Characteristics

 Demographic and HIV/HBV-related characteristics at cohort inclusion are described in Table 1. The median follow-up time was 37.1 months (interquartile range [IQR], 36.1-38.7). Previous anti-HBV treatment was taken in 146 (85.4%) patients, as the median duration prior to inclusion was 49.3 months (IQR, 25.8-66.4) for LAM and 4.9 months (IQR, 3.6-8.4) for ADV. The only anti-HBV treatments to significantly increase in use during follow-up were ADV and TDF (Fig. 1A). Other clinical features of HBV and HIV infection over time are summarized in Table 2.

Table 1. Study Population Characteristics (N = 171)
  1. a

    n = 150 ARV-experienced patients.

  2. b

    n = 161.

  3. Abbreviations: cART, combined antiretroviral therapy; F, female; M, male.

Age, years, mean (SD)40.0 (8.1)
Sex, M/F, no. (% male)154/17 (90.1)
Originating from zone of high HBV prevalenceb33 (19.3)
HIV characteristics 
HIV duration, years, mean (SD)9.6 (5.3)
Prior AIDS-defining illnes, no. (%)49 (28.7)
Duration of prior cART, years, mean (SD)a5.8 (3.0)
HBV characteristics 
Prior treatment, no. (%) 
LAM145 (84.8)
FTC1 (0.6)
Famciclovir3 (1.8)
ADV9 (5.3)
TDF3 (1.8)
Interferon/pegylated interferon7 (4.1)
Precore mutations (W28), no. (%)b47 (29.2)
HBV genotype, no. (%) 
A106 (62.0)
B1 (0.6)
D17 (9.9)
E19 (11.1)
G19 (11.1)
Mixed A/G or A/D9 (5.3)
Figure 1.

HBV-DNA and anti-HBV treatment during follow-up. (A) Nonparametric test for trend during follow-up: treatment with LAM/FTC (P = 0.2), ADV (P = 0.03), or TDF (P < 0.001); HBV-DNA >190 IU/mL (P < 0.001). Bars represent the proportion of patients treated with the various anti-HBV NAs; the solid line represents the percentage of patients with HBV-DNA >190 IU/mL. (B, C) Among the 14 patients with biopsies before and at the end of follow-up, there was a significant decline in both (B) median total intracellular HBV-DNA (P = 0.002) and (C) median cccDNA (P = 0.02). Open circles represent undetectable levels. In this subgroup, treatment with TDF was predominant (LAM+TDF, n = 6; TDF only, n = 2; LAM+TDF+interferon, n = 3; FTC+TDF, n = 1), and two patients had only LAM-containing combined antiretroviral therapy.

Table 2. Virological and Biochemical Changes During Follow-up
ParametersnBaselineEnd of Follow-upa
  1. Only HBeAg-positive patients were included for HBeAg loss and seroconversion.

  2. a

    Data from last follow-up visit among patients with at least two points of follow-up (n = 158); median follow-up time was 37.1 months (IQR, 36.1-38.7).

  3. b

    Number of patients with an event (cumulative probability at end of follow-up).

  4. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; HBeAg, hepatitis B e antigen.

HBV parameters   
HBV-DNA log10, IU/mL, median (IQR)1585.60 (3.58-6.88)1.78 (1.78-3.72)
HBV-DNA, >190 IU/mL, no. (%)158144 (91.1)67 (42.4)
ALT, IU/mL, median (IQR)15555 (32-98)39 (27-59)
AST, IU/mL, median (IQR)15247 (30-72)32 (27-46)
Hyaluronic acid, μg/L, median (IQR)15041 (24-64)25 (17-46)
HBeAg lossb12027 (38.1)
HBeAg seroconversionb12013 (22.4)
HBsAg lossb1581 (0.7)
HIV parameters   
CD4+ count, cells/mm3, median (IQR)141400 (264-545)435 (323-580)
HIV-RNA, copies/mL, median (IQR)1411.70 (1.70-3.66)1.70 (1.70-1.70)
HIV-RNA, <50 copies/mL, no. (%)14177 (54.6)110 (78.0)
CD4+ count, cells/mm3, median (IQR)17369 (297-527)431 (298-508)
HIV-RNA, copies/mL, median (IQR)174.40 (4.02-4.66)3.71 (1.70-4.39)
HIV-RNA, <50 copies/mL, median (IQR)171 (5.9)8 (47.1)

Evolution of HBV Mutations Over Time

Alkyl-Phosphonate–Associated pol Gene

At inclusion, there was a low prevalence of alkyl-phosphonate–associated mutations, which were found in two patients (details of specific mutations are given in Supporting Table 2). Prevalence of alkyl-phosphonate (Fig. 2A) mutations minimally increased from 1.7% at inclusion to 7.8% at M36 (cumulative incidence at the end of follow-up: 7.4/100 person-years), with the largest gain in aa changes at position rt181 (n = 6) followed by one rtA194T and one rtN236N/T mutation (Supporting Table 2).

Figure 2.

Progression of HBV pol and S gene mutations during follow-up. Markers indicate the point prevalence of (A) alkyl-phosphonate–associated pol gene, (B) l-nucleoside–associated pol gene/antiviral-associated S gene, and (C) immune-associated S gene mutations given at each cohort visit. *Number of patients having information on all positions within mutation class (after specified handling of missing data; see Supporting Fig. 1)

d-Cyclopentane–Associated pol Gene

At inclusion, two patients had a d-cyclopentane–-associated pol gene mutation (Table 3), giving a baseline prevalence of 1.7%. One incident d-cyclopentane–associated mutation was observed during follow-up and was highly mixed at position rtL180L/M+rtM204V/I+rtT184T/S (cumulative incidence at the end of follow-up: 2.4/100 person-years). Of note, aa changes at positions rt169, rt202, and rt250 were also examined, yet were never present during follow-up.

Table 3. HBV Mutations on the pol and S Gene During Follow-up
EndpointnbCumulative Incident Ratea
1 Year2 Year3 Year
  1. a

    Per 100 person-years.

  2. b

    Some patients only had one visit at cohort inclusion and therefore were not included in follow-up analysis.

  3. c

    Incident mutation was defined as any new mutation at any of the positions rt173, rt180, and rt204.

  4. The following exclusions were made dependent upon HBV mutations at inclusion:

  5. d


  6. e

    rtM204V, rtL180M+M204V/I, or rtV173L+rtL180M+rtM204V/I;

  7. f

    rtM204I, rtL180M+M204V/I, or rtV173L+rtL180M+rtM204V/I;

  8. g

    rtL180M+M204V/I or rtV173L+rtL180M+rtM204V/I.

Alkyl-phosphonate–associated pol gene1151.87.4
d-Cyclopentane–associated pol gene1150.02.4
l-Nucleoside–associated pol gene/antiviral-associated S genec130d5.414.017.5
rtM204I (sW196a/S/L)74e0.01.44.5
rtM204V (sI195M)73f0.02.84.2
rtV173L+rtL180M+rtM204V/I (sE164D+sI195M)130d2.35.55.5
Immune-associated S gene1491.32.06.4

l-Nucleoside–Associated pol Gene/Antiviral-Associated S Gene

When examining antiviral-associated S gene/l-nucleoside–associated pol gene mutation class as a whole, baseline prevalence was established at 56.8% and slowly increased to a point prevalence of 68.9% at M36 (Fig. 2B). Incident mutations occurred in 25 patients, two of whom developed two mutations over time (Supporting Table 2). At the end of follow-up, cumulative incidence was highest for conserved site changes at positions rtL180M+M204V/I (sI195M or sW196*/S/L) and rtV173L+rtL180M+rtM204V/I (sE164D+sI195M), while stabilizing after 2 years (Table 3). A univariate analysis showed a variety of risk factors associated with baseline mutations; however, only high HBV-endemic region, cumulative *LAM-duration, log10 HBV-DNA and alanine aminotransferase levels remained significant in the multivariable model (Table 4). During follow-up, incident mutations were significantly associated with current LAM-use and inversely associated with cumulative TDF duration (Table 4).

Table 4. Risk Factors for HBV S Gene Mutations at Inclusion and During Follow-up
 At InclusionDuring Follow-up
aOR[1] (95% CI)PaHR[1] (95% CI)P
  1. Models were adjusted for all risk factors displayed in the first column. Information on univariate analysis and model selection is provided in Supporting Table 1. Mutation endpoints were classed as either l-nucleoside-associated pol gene/antiviral-associated S gene (rtM204I [sW196*/S/L], rtM204V [sI195M], rtL180M+M204V/I, rtV173L+rtL180M+rtM204V/I [sE164D+sI195M]) or immune-associated S gene (sP120S/T, sG145A/K/R).

  2. a

    Four patients were excluded from analysis due to missing data on ALT.

  3. b

    Twenty-five incident mutations were considered, while defined as any new mutation at positions rt173, rt180, and rt204.

  4. c

    Only nine incident mutations were considered; one patient was excluded because of a mutation at position s145 at inclusion.

  5. Abbreviations: aHR, adjusted hazard ratio; ALT, alanine aminotransferase; aOR, adjusted odds ratio; NS, not significant.

l-Nucleoside–associated pol gene/antiviral-associated S gene
Originating from zone of high HBV prevalence0.32 (0.11-0.97)0.04NS 
Concomitant treatment with LAMNS 4.61 (1.36-15.56)0.01
Cumulative treatment with LAM (per month)1.03 (1.02-1.05)<0.001NS 
Cumulative treatment with TDF (per month)NS 0.94 (0.89-0.98)0.01
ALT (per 10 IU/mL)1.09 (1.00-1.19)0.047NS 
HBV-DNA (per log10 IU/mL)1.31 (1.07-1.61)0.009NS 
Immune-associated S gene
Age (per year)1.13 (1.03-1.23)0.01NS 
Originating from zone of high HBV prevalence12.95 (1.68-99.55)0.01NS 
Cumulative treatment with ADV (per month)1.23 (1.01-1.49)0.04NS 
Cumulative treatment with TDF (per month)NS 0.88 (0.79-0.98)0.02
HBV genotype D + A/D9.62 (1.16-79.85)0.04NS 

Immune-Associated S Gene

S gene mutations directly linked to immune-selective pressures had rather low prevalence at cohort inclusion, with the majority of conserved site changes at position s145 (Supporting Table 2). Prevalence of these mutation types slowly increased from 4.8% at baseline to 11.3% at M36 (Fig. 2C). In total, six patients developed HBV strains with aa changes at position s120 and four at position s145, for a cumulative incidence of 6.4/100 person-years at the end of follow-up (Table 3). In a risk factor analysis, patients originating from areas of high HBV prevalence, age, HBV genotype D, and cumulative duration of ADV treatment were significantly associated with baseline mutations (Table 4). With respect to incident mutations, host factors were no longer significantly associated and only cumulative TDF duration emerged as a significantly protective determinant (Table 4).

In a post hoc analysis, we decided to model CD4+ cell count changes as a risk factor for incident immune-associated S gene mutation. First, changes in CD4+ cell count from the previous annual visit were included in the model from Table 4 as a time-varying covariate; however, no significant association was observed (adjusted hazard ratio [HR]/100 mm3, 0.95; 95% confidence interval [CI], 0.75-1.20; P = 0.6). Second, we used more clinically relevant thresholds, including CD4+ cell count at 350/mm3 as a time-varying covariate. The rate of immune-associated S gene mutations was higher as patients spent more time above 350 cells/mm3 (adjusted HR, 1.46; 95% CI, 0.88-2.43), yet this was not significant (P = 0.14).

Treatment Strategies and HBV Mutations

 Because treatment stood out as a major predictor of mutation incidence, treatment strategies during follow-up were examined further. As shown in Table 5, incident antiviral-associated S gene/l-nucleoside–associated pol gene mutations occurred in the majority of patients with continuous LAM and in nearly a third of patients switching from LAM to ADV or multiple switches. Only four of the 25 patients harboring these incident mutations had undetectable HBV-DNA at the end of follow-up, either as the result of several treatment modifications or from simply LAM to ADV+LAM or TDF. Immune-associated S gene mutations mainly arose when patients were not undergoing treatment, switching from LAM to ADV, or continuing on LAM (Table 5). At the end of follow-up, all patients with incident immune-associated S gene mutations had detectable HBV-DNA without treatment modification/intensification. It should be noted that in the mutation classes above, no patient undergoing TDF while exhibiting virological rebound/nonresponse at the end of follow-up had detectable levels of TDF.

Table 5. Antiviral Treatment Strategies and HBV Mutations
Mutation TypeNo TreatmentNo Previous Anti-HBV NA ExposurePrevious Exposure with LAM Switching to:
LAMaLAM+TDFNo TreatmentADV±LAMTDF±LAM/FTCMultiple Switchesb
  1. All values are presented as no. (%). Table includes only patients with follow-up (n = 158). Mutation classes were defined in the following groups (aa changes): alkyl-phosphonate–associated pol gene (rtA181T/V/G, rtA194T, rtN236T); d-cyclopentane–associated pol gene (rtL180M+rtM204V+rtT184S); l-nucleoside–associated pol gene/antiviral-associated S gene (rtM204I [sW196*/S/L], rtM204V [sI195M], rtL180M+M204V/I, rtV173L+rtL180M+rtM204V/I [sE164D+sI195M]); immune-associated S gene (sP120S/T, sG145A/K/R).

  2. a

    The median duration of LAM prior to follow-up was 45.6 (IQR, 16.3-66.7) in this treatment group.

  3. b

    Of the 22 patients in this group, 16 and 4 patients ended follow-up with a TDF- and ADV-containing cART regimen, respectively. One patient switching from ADV to TDF+FTC also belonged to this group.

  4. c

    Patients with baseline mutations were not included in follow-up.

  5. d

    Only patients without rtV173L+L180M+M204V/I mutations at inclusion were included.

  6. Abbreviation: cART, combined antiretroviral therapy.

Alkyl-phosphonate–associated pol gene(n = 10)(n = 17)(n = 6)(n = 8)(n = 11)(n = 51)(n = 16)
Baseline000001 (2.0)1 (6.3)
Incidentc1 (11.1)2 (11.8)001 (9.1)3 (6.1)1 (6.3)
d-Cyclopentane-associated pol gene(n = 10)(n = 17)(n = 6)(n = 8)(n = 11)(n = 51)(n = 16)
Baseline01 (5.9)00001 (6.3)
Incidentc01 (6.3)00000
l-Nucleoside–associated pol gene/antiviral-associated S gene(n = 11)(n = 19)(n = 10)(n = 8)(n = 12)(n = 76)(n = 22)
Baseline2 (18.2)10 (52.6)03 (37.5)11 (91.7)51 (67.1)13 (59.1)
Incidentd1 (11.1)9 (52.9)1 (10.0)1 (12.5)2 (28.6)6 (9.4)5 (33.3)
Immune-associated S gene(n = 11)(n = 19)(n = 10)(n = 7)(n = 12)(n = 76)(n = 22)
Baseline00001 (8.3)3 (4.0)4 (18.2)
Incidentc3 (27.3)2 (10.5)002 (18.2)1 (1.4)1 (5.6)

Biochemical Fluctuation During Incident S Gene Mutations

 On a clinical note, alanine aminotransferase and aspartate aminotransferase levels 5 times the upper limit of normal were not frequent during incident mutation, occurring in 2/25 patients with antiviral-associated S gene/l-nucleoside-associated pol gene and 2/9 patients with immune-associated S gene mutations. When examining replication, aminotransferase levels, and hyaluronic acid levels, patients with incident mutations exhibited different longitudinal profiles (Supporting Fig. 2). No other major liver-related complications (e.g., fulminant hepatitis, decompensated liver, and HCC) were observed in patients with incident mutations.

Total Intracellular HBV-DNA, cccDNA, and HBV Mutations

 Using information from liver biopsies (n = 14, demographic and clinical data in Supporting Table 3), both total intracellular HBV-DNA and cccDNA decreased substantially during follow-up (Fig. 1B and Fig. 1C, respectively). Median total intracellular HBV-DNA dropped from 1.78 log10 copies/cell (IQR, 0.27-2.22) at inclusion to 0.18 (IQR, −0.39 to 0.30) at month 36 and median cccDNA decreased from −0.33 log10 copies/cell (IQR, −0.90 to 0.71) at inclusion to −1.46 (IQR, −2.00 to −0.58) at month 36. Among the five patients with HBV genetic information, virological and treatment data over time are reported in Table 6.

Table 6. Clinical and Virological Summary Among Patients With Liver Biopsies and Available HBV Genetic Information During Follow-up
Patient No.Anti-HBV TreatmentGenotypic Mutations in SerumSerum HBV-DNA (log10 IU/mL)Total Intracellular HBV-DNA (log10 Copies/Cell)cccDNA (log10 Copies/Cell)
InclusionDuring Follow-upaInclusionMonth 36InclusionMonth 36InclusionMonth 36
  1. a

    Only incident mutations are included.



 Using prospective data from 171 HIV/HBV-coinfected patients, this study reports incidence rates and determinants of a vast array of pol and S gene mutations under treatment-induced selective pressures. Single-sample prevalence for a number of these mutations has already been described in HIV/HBV-coinfected patients, showing strong agreement with those reported at inclusion in the present study.[16-19] While under close follow-up, an increase of incidence was observed in almost all mutation classes, especially l-nucleoside–associated pol gene/antiviral-associated S gene. The development of immune-associated mutations on the S gene, known to pose potentially serious public health problems,[20] approached over 10% of the study population by the end of follow-up.

The rtA194T mutation, previously described to be phenotypically associated with decreased sensitivity to TDF,[18] was observed in only one TDF-treated patient. This patient exhibited HBV viral load decline that was typical of TDF-treated coinfected patients over the first 6 months of therapy.[21] HBV-DNA virological response was achieved and sustained for a long period of time, confirming other in vitro data.[22] However, the concomitant mutation pattern in this patient, specifically on the precore and basal core promoter, may not have been exactly the same as described necessary for optimal TDF resistance.[23]

As reported in HBV monoinfection,[24] ETV resistance mutations were observed in three patients with LAM resistance, despite no exposure to ETV. These two mutation classes are highly linked. In vitro research has shown a very high increase in LAM resistance with the addition of ETV-resistant mutations.[25] A previous case study has also demonstrated the continual selection of LAM resistance mutations while under ETV therapy, resulting in a complex mixture of mutations similar to the patient with incident ETV resistance.[26] Taking the above into consideration, entecavir still needs to find its place among available anti-HBV treatments in HIV/HBV-coinfection, albeit preliminary results regarding its efficacy with concomitant TDF/FTC have been shown in patients with persistently low-level HBV replication.[27]

Knowledge of risk factors for both immune-associated and antiviral-associated mutations is presently scarce, with single-sample studies consistently reporting age and genotype D as determinants.[19, 28] We observed several HBV genotypes associated with incident S gene mutations, similar to a more recent study demonstrating mutations on the major hydrophilic region of the HBsAg across many genotypes.[29] Nevertheless, HBV genotype D was significantly associated with baseline immune-associated mutations, possibly owing to the natural increased genetic distance and higher diversity of the S gene observed for this particular genotype.[30] When comparing separate analyses for immune-associated S gene mutations at inclusion versus during follow-up, we found that many of the host-related or HBV-related factors were no longer significant and treatment surfaces as a major determinant. Cross-sectional results may primarily reflect the broader historical accumulation of events leading up to these mutations. For example, patients from a region of high prevalence, whose HBV transmission can be largely attributed to maternal-infant transmission, presented with higher odds of prevalent mutations, but no significant increase in hazards with incident mutations. It is worth mentioning that estimated duration of HBV infection was not the most accurate marker of determining this historical impact, since many of our patients were tested for HBsAg serostatus at the same time as anti-HIV antibodies.

The therapeutic implications of S gene mutation apparition have been thus far unclear for the HIV/HBV-coinfected population. LAM-exposure is an obvious risk factor for antiviral-associated S gene mutations, since it is directly involved with the overlapping LAM-resistant pol mutations. With immune-associated mutations, the link is less obvious. First, the presence of anti-HBs antibodies blocks entry of circulating HBV virions into the hepatocyte and possibly release of infectious particles from the infected hepatocyte,[31] during which time HBV mutants able to escape antibody recognition are being selected. By removing circulating HBV via antiviral therapy, this selective process could be mitigated. Perhaps selection of immune-associated mutations still occurs under less potent therapies, which is probably the case with LAM or ADV.[32] In our cohort, the majority of incident immune-associated mutations were observed in patients with previous or current LAM exposure, but rarely under TDF therapy, regardless of previous LAM exposure. In addition, other changes on the pol gene are known to arise during ineffective LAM therapy, resulting in higher variability in the overlapping S gene.[16, 33] Inefficient virological suppression during periods of low-potency antiviral therapy could therefore allow a larger pool of mutated virus from which immune-associated S gene mutations could be selected.

The incidence of immune-associated S gene mutations appeared to be slightly, however nonsignificantly, influenced by CD4+ cell count, where risk of mutation incidence was higher with increased levels of immunocompetence (determined by follow-up time >350/mm3). Amino acids on the “a” determinant have been found to be significantly less conserved in HBV monoinfection than HIV/HBV coinfection,[34] implying that higher levels of immune status are associated with increased selective pressures. This finding could be explained by the defect of specific anti-HBV immune response in coinfection, from processes such as anti-HBV CD8+ cytotoxic T cell exhaustion[35] or reduced T cell activation.[36]

It is fairly unknown whether mutations on the S gene directly translate to increased risk in clinical sequelae. Indeed, in a large cohort of HIV/HBV-coinfected patients, LAM-resistant HBV strains were observed in two of four patients who died from liver disease.[37] In the REVEAL study among HBV-monoinfected patients, a much stronger association with HCC was observed with persistently elevated HBV-DNA replication rather than other genetic information, such as HBV genotype (HR >8.0 versus 2.1, respectively).[38] The fact that none of our patients experienced any major liver complications and the majority had controlled HBV-DNA at the end of follow-up suggests that the mutations themselves may not be associated with any specific deleterious outcome, rather persistent HBV-DNA replication.

Another uncertainty is how intracellular DNA levels with mutations change during the course of TDF treatment. The two patients with LAM-resistant mutations had a similar 3-year decline in total intrahepatic and cccDNA levels compared with patients without mutations. Moreover, the patient with ADV resistance exhibited low levels of total intrahepatic and cccDNA at the beginning and end of follow-up. These findings suggest that replication of these HBV mutants can also be affected by potent antiviral therapy at the intracellular level. Interestingly, LAM- and ADV-resistant strains have been isolated from hepatocytes when not detected in the periphery.[39] Unfortunately, no further conclusions regarding the changes in quasispecies makeup within the hepatocyte can be provided, as this data was not obtained.

Certain limitations need to be addressed in our study. Not all S gene mutations associated with decreased antigenicity were detected in this patient population, particularly between positions 121 and 123.[6] Because aa changes at only positions 120 and 145 were determined, underestimation of the immune-associated mutations is possible. Second, some caution should be used when interpreting analyses with small numbers of S gene mutations. The lack in power could have masked some potential risk factors. Furthermore, certain multivariable models could have been overfit, such as in the model with immune-associated mutations at inclusion where rather large confidence intervals were observed. Third, our study population had a mix of various antiviral regimens prior to inclusion, with the majority being LAM-experienced. The selective pressures induced during this period could have increased the probability of changes in the HBV genome, thereby rendering the population more at risk of developing certain mutations. Finally, treatment adherence was only monitored on a selected group of patients and not the entire cohort.

In conclusion, there is a substantial increase in a number of HBV pol and S gene mutations among HIV/HBV-coinfected patients over time. The emergence of these mutations is closely linked to antiviral-selective pressures. This finding can be particularly worrisome in settings such as Sub-Saharan Africa and Asia, where there is a constrained access to powerful anti-HBV drugs and virological monitoring and a high prevalence of HBV infection. For the time being, the use of more potent antivirals with a high barrier to resistance and close virological follow-up are highly recommended in coinfected patients in order to reduce the probability of more problematic HBV mutations emerging. In HIV infection, emerging S gene mutations might jeopardize immunization and prevention strategies aimed at reducing transmission of HBV mutant strains.


We thank H. Rougier, L. Roguet, and M. Sébire for managing the logistics of the French HIV-HBV Cohort and G. Pannetier and F. Carrat for help with data management. We are also grateful to the patients and clinical teams for their commitment to the French HIV-HBV Cohort Study.