Ten Years Survey of Primary HIV-1 Resistance in Serbia: The Occurrence of Multiclass Resistance

Abstract

In Serbia, the first cases of HIV infection were reported in 1985, whereas antiretroviral (ARV) therapy has been in use since 1987. With this study we aimed to assess the occurrence and pattern of HIV resistance mutations among newly diagnosed patients in the period 2002–2011. The study prospectively included 181 adult patients. Genotypic HIV-1 drug resistance testing was performed and drug resistance was scored according to the 2009 WHO list for surveillance of drug resistance mutations (SDRMs). A bioinformatic approach was used to estimate the duration of infection by calculating the percentage of ambiguous basecalls per sequence, with a cutoff of 0.47% as the delimiter for recent infection. The overall prevalence of transmitted drug resistance (TDR) found in the study was 8.8% (16/181, 95% CI=5.5–13.8). Thirty-one percent of resistant samples contained multiple SDRMs. In particular, 5/16 patients with resistance carried viral strains with SDRMs to multiple ARV classes, hence one-third of resistant strains were multiclass resistant, including non-B strains. A total of 51.9% of samples (94/181) were classified as recent infection, with a significant increase in the second part of the study period. However, the prevalence of TDR in recent infection was 6.4% (6/94, 95% CI=2.9–13.2), not statistically different from that found in nonrecent infection. We showed a changing pattern of TDR mutations over the study period, with a substantial occurrence of multiclass resistance, across different HIV subtypes. Our results highlight the need for continued surveillance of primary resistance.

Introduction

The advent of highly active antiretroviral therapy (HAART) has significantly reduced morbidity and mortality among HIV-infected patients.¹ However, successful treatment is not always achieved and may be impaired by the development of HIV drug resistance, being both an important cause as well as a consequence of therapy failure.^2–4 A particular problem is the occurrence of transmitted drug resistance (TDR) among treatment-naive patients. Growing literature data on the rate of TDR give various estimates as to the prevalence of TDR in different geographic settings, ranging from 0% to 52%.^4–8 Still, direct comparison is often not feasible, due to different interpretation criteria and algorithms used to assess TDR. Published pan-European surveillance studies, which systematically gathered data from a large number of European countries, reported the mean prevalence of TDR to be rather constant, around 9–10%, through almost a decade, from 1996 to 2005.^9–11 Onward transmission of resistance mutations among newly infected persons may present a significant public health problem.¹² The estimation of the rates and prevailing patterns of TDR is crucial for HIV surveillance programs, providing feedback on the efficacy of prevention and having implications to first line treatment strategies.

The first cases of HIV infection in Serbia were reported in 1985.¹³ From the beginning of the epidemic until the end of December 2012, 2,850 cases of HIV-1 infection were registered, of which 1,645 (57.7%) were AIDS cases.¹⁴ The registered incidence of HIV infection in 2011 was 17 per million, in the range of 100–150 new cases per year.¹⁵ According to the latest epidemiologic data, among the newly diagnosed patients infection via sexual transmission is the most prevalent, accounting for over 80% of newly registered cases in 2012.¹⁴ Regarding HIV subtype distribution, a recent study has confirmed previous findings about the predominance of subtype B (over 90%), with other subtypes also present.^16,17 In Serbia, antiretroviral therapy (ART) was introduced in 1987. A number of factors potentially facilitating wider development of HIV drug resistance existed in Serbia: periodic limited access to laboratory monitoring of viral load and drug resistance testing leading to prolonged exposure to probably failing regimens, as well as limited options for sequential antiviral therapy, if required.

To date, only one study on HIV resistance in Serbia has been reported¹⁸ and no detailed analysis of the occurrence and patterns of mutations related to primary HIV resistance in Serbia has been done. With this study we aimed to assess the occurrence of primary resistance and patterns of resistance mutations among newly diagnosed HIV-positive patients in Serbia in the period 2002–2011.

Materials and Methods

The study prospectively included newly diagnosed patients aged 18 years or more, in the Center for HIV/AIDS, Institute for Infectious and Tropical Diseases in Belgrade, from September 2002 to December 2011, with informed consent from the patients, partly within the European projects for monitoring of primary HIV resistance SPREAD/EuropeHIVResistence. A total of 583 samples were collected in the study period, of which 181 patients were randomly selected to be included in the analysis. Included patients had no prior exposure to ART and their viral load was determined to be higher than 1,000 c/ml (log 3). Epidemiologic, clinical, and behavioral data were collected using a standardized questionnaire. Genotypic resistance testing was done on a sample taken within 6 months of diagnosis, by using commercial and/or an in-house genotypic resistance assay: Viroseq HIV-1 Genotyping System (Celera Diagnostics, Alameda, CA) (53 samples), Trugene HIV Genotyping Kit (Visible Genetics, Toronto, Canada) (32 samples), according to the manufacturer's recommendations, and an in-house genotypic resistance assay (96 samples).

For in-house processed samples, plasma samples obtained from EDTA whole blood were stored at −80°C prior to genotypic testing. A total of 1 ml of each plasma sample was centrifuged for 1 h at 4°C at 20,000 rpm. The supernatant was carefully removed to the volume of 280 μl; the pellet was resuspended and used for RNA extraction. RNA was extracted using a QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol and was subjected to genotyping by an in-house nested polymerase chain reaction (PCR) protocol amplifying the HIV-1 pol gene.¹⁹ Reverse transcription was performed using the One Step RNA PCR Kit (Qiagen, Hilden, Germany), followed by a nested PCR protocol using the Taq PCR Core Kit (Qiagen, Hilden, Germany). The amplified products, 1.6 kb in length (full-length protease and near complete reverse transcriptase), were purified using the Qiagen PCR Purification Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol and sequenced bidirectionally on an ABI Prism 310- Genetic Analyzer (Applied Biosystems, Foster City, CA) using Big Dye Terminator chemistry and six sequencing primers.¹⁹ Obtained sequences were visually inspected and manually edited and then assembled with a SeqScape HIV-1 Genotyping System Software v. 2.5 (Applied Biosystems, Foster City, CA).

Mutations associated with TDR were scored according to a surveillance drug resistance mutation (SDRM) list by WHO²⁰ by using the Calibrated Population Resistance Tool version 6.0.²¹ HIV subtyping was performed using the REGA HIV-1 Automated Subtyping Tool (version 2.0).²²

A bioinformatic approach was used to estimate the duration of infection by calculating the fraction of ambiguous nucleotides in the sequence as a delimiter for more recent (less than 1 year) versus chronic infection (longer than 1 year).²³ Ambiguous mutations, representing a mixed nucleotide signal, were identified when the sequencing signal intensity of the minor base in both directions was ≥20% of the major base signal at a particular position. The method was applied to both the complete dataset and subtype B sequences only, also using three different ambiguity cutoff values (0.45%, 0.47%, and 0.5%) for recent infection.

The results were analyzed by standard statistical analysis. Prevalence values were calculated with a 95% Wilson score confidence interval based on a binomial distribution. Categorical data were compared using the chi-square test and Fisher's exact test. For statistical analysis we grouped the patients by the calendar year of diagnosis into two periods: 2002–2006 and 2007–2011, thus forming two patient groups of comparable size (95 and 86 patients, respectively).

Results

Study population

DNA sequencing was fully successful in 179 samples yielding full-length protease (PR) and more than 250 reverse transcriptase (RT) codons, whereas in two samples partial PR or RT regions were obtained. The baseline characteristics of the patients are summarized in Table 1.

Table 1.

Clinical and Demographic Characteristics of Patients

	Total 2002–2011 [number (%)]	2002–2006 [number (%)]	2007–2011 [number (%)]	p value
Sex
Female	29 (16)	20 (21)	9 (10)
Male	152 (84)	75 (79)	77 (90)
Total	181 (100)	95 (100)	86 (100)
Age
Mean (range)	31.4 (18–73)	29.6 (18–67)	33.2 (19–73)
Transmission route
MSM	96 (53)	35 (37)	61 (71)	p<0.0001
Heterosexual	55 (30)	38 (40)	17 (20)	p=0.0123
IVDU	9 (5)	8 (8)	1 (1)	p=0.0365
Blood product	3 (2)	3 (3)	0
Unknown	18 (10)	11 (12)	7 (8)
Total	181 (100)	95 (100)	86 (100)
CDC stage
CDC stage A	75 (41)	29 (30)	46 (53)	p=0.0022
CDC stage B	30 (17)	18 (19)	12 (14)
CDC stage C	61 (34)	36 (38)	25 (29)
Unknown	15 (8)	12 (13)	3 (4)
Total	181 (100)	95 (100)	86 (100)
CD4 cells/mm3
Mean (range)	261 (1–1,320)	180 (1–773)	325 (4–1,320)	p=0.0004
STI
Positive	45 (25)	14 (15)	31 (36)	p<0.001
Negative	118 (65)	68 (72)	50 (59)
Unknown	18 (10)	13 (13)	5 (5)
Total	181 (100)	95 (100)	86 (100)
HBV infection
	17 (9.4)	5 (5.3)	12 (13.9)
HCV infection
	12 (6.6)	9 (9.5)	3 (3.5)

MSM, men who have sex with men; IDU, intravenous drug user; hepatitis B virus (HBV) coinfection was determined by the presence of HBsAg; hepatitis C virus (HCV) coinfection was determined by the presence of anti-HCV antibodies; STI, history of sexually transmitted infections, including positivity to anti-HBcAg.

The majority of patients in the study were infected through sexual contact (83%). However, the prevalence of transmission by men having sex with men (MSM) was much higher in the second half of the study period (37% in 2002–2006 compared to 71% in 2007–2011, p<0.0001), whereas the prevalence of heterosexual transmission and the risk of transmission by intravenous drug use declined (p=0.0123 and 0.0365, respectively).

Thirty-four percent of patients presented at CDC clinical stage C, and this was significantly more common among older compared to younger patients (70% in the age group 41–50 years compared to 19% in the age group 20–30 years, p=0.0038) and patients with a lower education level (61% patients with a primary education compared to 34% patients with a secondary or higher education, p=0.0226). However, the percentage of diagnosis in the earlier disease stage (CDC stage A) tended to increase in the second half of the study period and this increase was shown to be statistically significant (30% vs. 53%, p=0.0022).

The prevalence of sexually transmitted diseases (STDs), taking into account hepatitis B virus (HBV) as a sexually transmitted infection, was significantly increased in the second half of the study period (36% vs. 15%, p<0.001). There was a significant difference in the prevalence of STDs among MSMs compared to patients infected heterosexually (31% versus 11%, p=0.0287).

HIV subtyping

We found that 167/181 sequenced strains (92.3%) were of HIV subtype B, 4/181 samples were characterized as subtype G (2.2%), the same as for subtype C (2.2%), whereas 2/181 samples (1.1%) were found to be of subtype A and 1/181 of subtype F (0.5%). Circulating recombinant forms (CRFs) were found in 3 of 181 patients (1.6%): CRF01_AE was found in two samples (1.1%), while CRF02_AE was found in a single patient (0.5%).

Prevalence and patterns of resistance

The overall prevalence of TDR found in the study was 8.8% (16/181, 95% CI=5.5–13.8). The frequency of TDR mutations to different antiretroviral (ARV) classes was 5.5% (95% CI=3.0–9.8) for nucleoside reverse transcriptase inhibitors (NRTIs), 4.9% for nonnucleoside reverse transcriptase inhibitors (NNRTIs) (95% CI=2.2–8.4), and 2.2% for protease inhibitors (PIs) (95% CI=0.9–5.5). In the first half of the study period (2002–2006) the frequency of SDRMs was 7.4%, compared to 10.5% in the subsequent period (2007–2011); however, this difference did not reach statistical significance. Regarding different ARV classes, the frequency of NRTI-related SDRMs remained very similar in the two periods (5.3% in 2002–2006 versus 5.8% in 2007–2011); the frequency of NNRTI-related SDRMs increased (2.1% in 2002–2006 versus 8.1% in 2007–2011), whereas the frequency of PI-related SDRMs decreased (3.2% in 2002–2006 versus 1.2% in 2007–2011); however, these differences did not reach statistical significance.

Among the mutations and polymorphisms detected were 20 mutations from the SDRM list, with 13 mutations on the NRTI agents, four TDR mutations on the NNRTI agents, and seven mutations on the PI agents. Among patients with evidence of drug resistance, 10/16 (62.5%) displayed drug resistance mutations against NRTIs, similar to the proportion of patients with drug resistance mutations against NNRTIs (9/16, 56.3%), whereas drug resistance mutations against PIs were seen in 18.8% of patients (3/16). The most frequent mutations associated with NRTI resistance were thymidine analog mutations (TAMs) at RT positions 210, 215, and 219, for a total of 43.7% of resistant samples. The single most common NRTI resistance position was RT215 (T215ISY), present in 25% of all resistant samples. Of note, this mutation was seen only in samples from the first half of the study period (2002–2006), when 57.1% of resistant samples contained it (p=0.0192). The most common NNRTI mutation was K103N, present in 37.5% of resistant samples in total. This mutation was seen only in samples from the second half of the study period (2007–2011), and the difference compared to the previous one was found to be highly statistically significant (p=0.0032). The most prevalent PI mutations were at positions 46 and 82, with a frequency of 12.5% of resistant samples each. A detailed pattern of SDRM mutations found in our study is shown in Table 2.

Table 2.

Detailed Pattern of Mutations from the WHO List for Surveillance of Transmitted Drug Resistance Found in Our Study

Sequences with SDRMs	NRTIs		NNRTIs		PIs		Subtype
1		T215IS	None		None		B
2		M41L, L74I, V75T, T215Y, K219N		K101E, Y181C, G190A		M46I, I84V, L90M	B
3	None		None			V82L	B
4		T215I	None		None		B
5		K65R, Q151M		Y181C	None		G
6	None		None			M46L	B
7		T215S	None		None		B
8		L74I, Y115F, M184V		K103N		L23I, M46I, I54V, V82F, L90M	F
9		K219Q		K103N	None		B
10		L210W	None		None		B
11		L210W	None		None		B
12	None			K103N	None		B
13	None			K101E	None		B
14	None			K103N	None		B
15	None			K103N	None		B
16		K65R, M184V		K103N, Y181C	None		B

Samples with multiclass resistance are in bold.

SDRMs, mutations for surveillance of transmitted drug resistance; NRTIs, nucleoside reverse transcriptase inhibitors; NNRTIs, nonnucleoside reverse transcriptase inhibitors; PIs, protease inhibitors.

Sixty-nine percent of samples displaying TDR (11/16) contained a single mutation from the SDRM list, of which 5/16 samples displayed only NRTI mutations (31%), 4/16 patients only NNRTI mutations (25%), and two patients only PI drug resistance mutations (12.5%). Thus 31% of resistant samples had two or more (up to nine) SDRMs to two or three ARV classes. Of note, 5/181 patients carried viral strains with SRDM to multiple ARV classes (2.8%, 95% CI=1.2–6.3); thus 5/16 strains displaying TDR (31.2%) manifested with multiple SDRMs conferring resistance to multiple ARV classes. In particular, two of five multiclass resistant strains were of the non-B subtype (one subtype G and one subtype F). The proportion of non-B subtypes among SDRMs containing samples was less than 10%, similar to the overall subtype distribution in the population studied.

Based on the threshold of 0.47% ambiguous bases per sequence, a total of 51.9% of samples (94/181) were classified as a recent infection, of a duration of less than 1 year, whereas among subtype B samples this percentage was 52.7% (88/167). Lowering the threshold to 0.45% did not influence the result, nor did raising it to 0.5% (except in the subtype B dataset, where two additional samples were identified as recent infection with the least stringent cutoff of 0.5%). In the first half of the study period (2002–2006) the percentage of recent infection was 36.8% (35/95) in the whole dataset and 36.4% (32/88) in subtype B samples only, while in the second part (2007–2011) this percentage was found to be 68.6% (59/86) and 70.9% (56/79), respectively. In both datasets analyzed (including non-B subtypes and subtype B samples only) this difference between the first and the second part of the study period was found to be statistically highly significant at an equal level (p=0.0003). Moreover, a comparison of the mean CD4 cell count between recently infected patients, based on a low percentage of sequence ambiguity, and those designated as patients with established infection, with the percentage of sequence ambiguity above the cutoff level, revealed a significantly higher CD4 cell count in the former group (p=0.0022). The overall prevalence of TDR in recent infection was 6.4% (6/94, 95% CI=2.9–13.2), not statistically different from the prevalence in nonrecent infection. The total number of sequences with a low percentage of ambiguous bases and also containing SDRMs was 6/16 (37.5%), equally distributed in the first and second half of the study period, with three samples each (three of seven in 2002–2006 and three of nine in 2007–2011).

Discussion

Here, we present the first comprehensive study of the prevalence and patterns of primary resistance in newly diagnosed HIV-infected patients in Serbia, a country with a low prevalence of HIV infection (less that 0.1%). The incidence rate of HIV-1 infection was 17 per million in 2011, ranging from 14 to 20 per million during the years of the study period (2001–2011).¹⁵ ART has been available in Serbia since 1987, starting with zidovudine monotherapy, whereas HAART was introduced in 1997. However, decades of ART usage in Serbia have coincided with a period of transitional economic, social, and political instability, resulting in an episodic irregular drug supply. Periodic stock-outs have also occurred, prohibiting essential laboratory monitoring. Furthermore, the available repertoire of ARV drugs, fully covered by the National health insurance system, is rather limited, with no new drugs included since 2003.²⁴ Thus, the therapeutic choices for HIV treatment that have been available in Serbia include zidovudine, didanosine, stavudine, and abacavir among the NRTIs; nevirapine and efavirenz among the NNRTIs; and saquinavir, nelfinavir, indinavir, fosamprenavir, lopinavir/ritonavir, and ritonavir for boosting therapy among the PIs. Nelfinavir and indinavir were withdrawn from clinical use in 2008, while stavudine was withdrawn in 2011. Enfuvirtide was introduced in 2007 for salvage treatment. Newer drugs, such as, tipranavir, atazanavir, etravirine, raltegravir, maraviroc, and even tenofovir/emtricitabine, although licensed worldwide since 2004, have not been used in Serbia. The current clinical practice in HIV treatment in Serbia has shown considerable success;^25,26 however, we speculate that this particular and rather unique history of ARV therapy might have influenced the level and pattern of transmitted drug resistance.

The present study included around 50% of all newly diagnosed patients in the 10-year study period. The demographics of the study population were comparable to the national HIV-1 epidemic, characterized predominantly by sexual transmission of infection (80%) and a male-to-female ratio of 7:1, according to the latest data.^14,15 As previously described for Serbia, subtype B remained the most prevalent HIV subtype.^16,17 Epidemiologic data indicate a shift in prevailing transmission risk toward predominantly sexual transmission, in particular MSM transmission, that is also seen in our study.

Transmitted HIV resistance is an important public health issue, with the potential to limit future therapeutic options, yet studies have documented the effectiveness of genotypic tests for drug resistance-naive patients as soon as TDR reaches 1%.²⁷ Current guidelines recommend resistance testing for all treatment-naive patients prior to the initiation of HAART. Our study was initiated within the European projects of monitoring primary HIV resistance— SPREAD/EuropeanHIVResistance, which have shown that, on average, one in 10 newly identified HIV-infected patients in Europe carries the virus containing resistance-associated mutations.^9–11 The overall TDR prevalence found in our study of 8.8% is in line with the estimated average TDR frequency in Europe. Similar studies performed in the neighboring region, or in countries with socioeconomic similarities to Serbia, described varying prevalences of primary resistance.

A recent study from Croatia, analyzing samples from 2006 to 2008, revealed a much higher prevalence of TDR of 22%, mainly related to the contribution of the phylogenetically important cluster of MSM carrying the T215S mutation.²⁸ Similarly, a rather high TDR prevalence of 16.6% was detected in a study performed in Hungary in a similar time frame (2008–2010).²⁹ However, no TDR was reported in a study from Montenegro,³⁰ although with a limited number of only 10 samples, as well as in a study from Cyprus, representing 88% of the known-living HIV-1-infected population in the country.³¹ A rather low prevalence of primary resistance of 2/66 (3%) was found in a survey from Albania, a somewhat higher prevalence of 3.9% was described in Georgia, and, similarly, 4.7% was found in a recent study from Slovenia.^32–34 Other countries, e.g., Bulgaria and Poland, have reported an overall prevalence of TDR in the same range as the European average—9.1% and 7.4%, respectively.^35,36

Studies on the changes in the prevalence of TDR over time in different European countries have reported conflicting results, from a decline in prevalence, over stability, to an increase in TDR prevalence, reflecting differences in patient populations and practices of the antiretroviral regimen used.^37–40 Our study revealed an increase in overall SDRM frequency in the second half of the study period compared to the first half (from 7.4% to 10.5%); however, this difference did not reach statistical significance. This increase was mainly due to the increased frequency of NNRTI-related mutations (from 2.1% to 8.1%). In particular, the K103N mutation appears as TDR only in the second part of the study period, after 2007. For clinical practice, NNRTI resistance mutations are of particular interest since these mutations generally confer high-level cross-resistance to this drug class, in particular to the first generation of NRTIs that is available in Serbia, otherwise an important portion of many first-line regimens. Increased rates of the K103N mutation in untreated patients in Serbia may also be linked to the fact that this mutation does not substantially affect viral fitness and may persist for a prolonged period of time after transmission, with the possibility of being transmitted to others.⁴¹

Similar to other reports, the most frequent mutations associated with NRTI resistance were TAMs, associated with resistance to zidovudine and stavudine, probably reflecting the extensive use of this drug class in early treatments, even as monotherapy.⁴² Interestingly, however, TAMs at position RT215, including T215 revertant mutations (T215YIS), present in nearly 60% of resistant samples from the first half of the study period (2002–2006), virtually disappeared from our sample after 2007. This change in mutation pattern may have resulted from changes in prescribing practices: more extensive use of abacavir and lamivudine as the NRTI backbone.

In agreement with literature data,⁸ a majority of patients with TDR in our study had singleton resistance mutations, mostly to NRTIs (31%), to a lesser extent to NNRTIs (25%), and the least frequently to PIs (12.5%). Multiple SDRMs, conferring resistance to multiple ARV classes, including multidrug resistance (MDR) mutations, were present in 2.8% of the total number of samples; hence over 30% of resistant strains contained SDRMs to multiple ARV classes. The majority of studies in Europe have found only limited representation of multiclass resistance in naive patients so far, with the exception of a recent study from northern Greece that reported a prevalence of 11.8%.⁴³ A recent study from the United States has reported a substantial prevalence of 7.7% of multiclass resistance in a young cohort of naive patients and also identified an MSM transmission route as a risk for multidrug resistance.⁴⁴ The overall prevalence of TDR found in our study is similar to the prevalence described in Europe; however, the proportion of multiclass resistance is substantially higher, with almost one-third of resistant strains characterized by multi class resistance. In particular, two of five multiclass resistant strains were of subtypes G and F, respectively.

A number of studies have described a lower prevalence of TDR, to any of the drug classes, in patients infected with non-B viruses. In the absence of drug-induced selective pressure, the natural polymorphism in non-B subtypes may result in mutations at resistance-associated positions, with no relation to transmission of resistance.⁴⁵ Conversely, multiclass resistant non-B subtypes described in our study are characterized by a rather elaborate mutation pattern: the subtype F sample contained a triple class resistance mutation pattern (L74I, Y115F, M184V, K103N, L23I, M46I, I54V, V82F, L90M), whereas the subtype G sample contained multidrug resistance (MDR) mutations involving K65R and Q151M, together with Y181C. A recent study from Italy reported a remarkable frequency of TDR in the F1 subtype-infected patients of 15.4%, albeit mostly to a single class.⁴⁶ The substitution of glutamine by methionine at RT position 151 (Q151M) represents the primary mutation of the multinucleoside resistance complex. By itself, this mutation confers intermediate to high level resistance to multiple NRTIs. To date, there are very few reports of the occurrence of Q151M as a transmitted mutation without prior exposure to NRTIs.^42,47,48

Many drug resistance mutations incur a substantial fitness cost to the resistant virus; thus their replacement with wild-type variants potentially confers a survival advantage. Upon transmission of a drug-resistant strain, the emergence of a wild-type virus results from the occurrence of back-mutations, rather than the predominance of an existing wild-type variant. A number of studies have reported the differing persistence of TDR mutations.^41,48,49 Estimates of the level of transmitted HIV drug resistance may be affected by the average time elapsed between HIV-1 infection and genotyping. To address this possible limitation in our study, we have used a recently devised bioinformatic approach to assess the duration of an HIV infection, based on the percentage of ambiguous base calls in the sequence.

HIV-1 infection is associated with a transmission bottleneck: newly acquired HIV infection (in particular when sexually transmitted—which is the case in almost 90% of our study population) is established by a limited number or a single viral strain.⁵⁰ Further intrapatient viral evolution over time leads to increasing HIV viral diversity within individuals by an initial accumulation of mixed bases, as new mutations emerge in existing quasispecies. Hence, a threshold of the percentage of ambiguous bases per sequence has been proposed to identify nonrecent versus recent HIV infection (one lasting less than 6 months to 1 year).^23,51 Essential for this method's performance is standardized mixed bases calling, a variability that lays the ground for intrasample and intersample interpretation, using either automated or manual sequence editing. However, it has been shown that the duration of HIV infection can be inferred from the proportion of mixed bases identified during population-based sequencing of the pol region, with a similar accuracy using any mixed base threshold between 15% and 25%.⁵²

The threshold we used falls within the proposed range. Furthermore, a cutoff of 0.45–0.5% of ambiguous nucleotides for distinguishing recent from established infection has been proposed for HIV subtype B pol gene sequences, whereas a differing percentage has been described for some of the non-B subtypes.⁵³

Using this method we have consistently identified an increasing proportion of recent infections, significantly higher in the second half of the study period, using a different percentage ambiguity cutoff (0.45%, 0.47%, 0.5%) and also on both a complete and partial (only subtype B) dataset. This finding is in accordance with the decreasing proportion of late presenters, based on the CDC clinical stage, in the same time frame. Classification as recent infection, based on a low percentage of sequence ambiguity, is further supported by the fact that patients in the recent infection group had a significantly higher CD4 cell count compared to those designated as established infection, implying a shorter duration of infection. However, we found no difference in the prevalence of transmitted resistance between the two groups, implying that the occurrence of TDR is stably maintained in the population. However, studies based on dense sampling would be needed to further explore this phenomenon, possibly also including other approaches to assess the duration of HIV infection, such as avidity testing, “detuned” ELISA, or other serology-based tests.⁵²

Notably, we found that a lower education level was associated with late presentation. In other studies educational level has been shown to be associated with an increased risk of TDR and a poorer virologic and immunologic response to treatment.³⁷ An additional shortcoming of our study could have resulted from the population sequencing approach for genotyping, with a detection threshold of approximately 15% to 25%, limiting the sensitivity for minor resistance variants. This may have contributed to underestimating the prevalence of transmitted HIV drug resistance, and may have also biased the type and pattern of detected mutations, preferentially limiting the sensitivity for those that rapidly become undetectable such as K65R or M184V.³⁸

In conclusion, transmitted HIV drug resistance was found in around 9% of newly diagnosed patients in Serbia in the 10-year period 2002–2011, similar to the European average. We showed a changing pattern of TDR mutations over the study period, with a substantial occurrence of multiclass resistance across different HIV subtypes. Our results highlight the need for continued and improved surveillance of primary resistance.

Sequence Data

Sequences analyzed in this study are available at the NCBI Nucleotide Sequence Database under the following accession numbers: GQ399763.1, GQ399551.1, GQ400327.1, GQ400482.1, GQ398972.1, GQ399179.1, GQ399012.1, GQ399888.1, GQ399221.1, GQ399018.1, GQ480327.1, GQ399263.1, GQ395533.1, GQ400459.1, GQ400490.1, GQ400092.1, GQ399955.1, GQ400505.1, GQ399341.1, GQ399810.1, GQ398855.1, GQ400529.1, GQ400303.1, GQ400203.1, GQ400169.1, GQ399328.1, GQ398698.1, GQ399770.1, GQ399505.1, GQ399605.1, GQ399463.1, GQ399262.1, GQ400380.1, GQ399684.1, GQ399526.1, GQ400192.1, GQ399335.1, GQ400562.1, GQ400943.1, GQ400568.1, GQ400867.1, GQ399293.1, GQ400985.1, GQ400636.1, GQ400934.1, GQ400637.1, GQ400971.1, GQ400727.1, GQ400860.1, GQ400842.1, GQ400711.1, GQ400847.1, GQ400975.1, GQ401005.1, GQ400863.1, GQ399151.1, GQ400634.1, GQ400696.1, GQ400623.1, GQ400576.1, GQ400698.1, GQ400683.1, GQ400664.1, JX299860.1, JX300595.1, JX300466.1, JX301157.1, JX300670.1, JX300934.1, JX300963.1, JX301026.1, JX299883.1, JX299967.1, JX300342.1, JX300698.1, JX301113.1, JX299941.1, JX300732.1, KF056325, KF157531, KF362128, KF157408–KF157507.

Acknowledgments

The work was funded by the Ministry of Education and Science Republic of Serbia, grant 175024, and partly funded by FP6 EC grant LSHP-CT-2006-518211. We thank Anne-Mieke Vandamme and Dimitrios Paraskevis for providing genotyping analysis in the study period 2002–2004.

Author Disclosure Statement

No competing financial interests exist.