HIV-1 subtype diversity is the highest in the world. Over nine pure subtypes of HIV-1 group M are currently known (A-D, F-H, J and K). Some of these subtypes are further subdivided into sub-subtypes e.g. subtypes F (F1 and F2); A (A1, A2 and A3). Intrasubtype divergence of up to 20% also occurs. Furthermore, intersubtype divergence of about 25% to 35% occurs, for the env amino acid sequences [1] [4] [5] . This intermixture of HIV-1 variants that circulate together within a geographical region provides an opportunity for recombination of virus strains within dually or multiply infected individuals [6] . Some of these recombinant forms may further achieve epidemic relevance, giving rise to unknown circulating recombinant forms (CRFs) thus complicating therapy. To date, over 40 CRFs are recognized in diverse parts of the world [7] . It is currently believed that HIV-1 M subtypes and CRFs are the result of founder effects in different geographic locales, followed by localised evolution. As a consequence, such HIV-1 forms are heterogeneously spread out worldwide [2] .

Africa has all HIV-1 subtypes, although A and C predominate [8] . In East Africa, subtypes A and D have dominated Uganda from mid 1980s [9] . These subtypes A and D have also been isolated from various parts of Tanzania [10] . However, in the Southern Tanzania town of Mbeya, the predominant subtype is C [11] . This is believed to have been introduced from Southern African countries where subtype C is predominant [12] .

HIV-1 epidemic in Kenya like the rest of the world shows regional heterogeneity. On the basis of the env C2-V3, Neilson et al., 1999; found that subtype A predominates (71% - 87%), with significant components of subtype D (7% - 29%) and subtype C (7% - 17%). Carr et al., 2005 [14] has identified a full-length subtype G in Kenya. On the other hand, it has been established that HIV-1 subtypes circulating in the Northern Kenya town of Moyale mainly comprise subtype A (50%), subtype C (39%) and subtype D (11%); i.e. subtypes A and C are predominant. This is probably influenced by neighbouring Ethiopia which is dominated by HIV-1 subtype C. It has been suggested that this scenario is indicative of cross-border movements influencing circulation of subtypes in northern Kenya [13] . Evidence of increasing HIV-1 diversity is emerging. For example, Lihana et al., 2006 [15] identified four circulating recombinant forms: between A1, A2, and D; A2 and D; A1 and D; A1 and G; A1 and C; A1, C and D have been documented among STI patients in Nairobi.

In the context of this increasing HIV-1 subtype diversity, it requires further studies in other parts of Kenya to monitor for mutations that are likely to lead to drug resistance if we have to achieve successful treatment outcomes. Thus, this study was conducted in western Kenya bordering Uganda.

The study was conducted in western Kenya at sites within the border area of Kenya and Uganda. The five sites: Butula, Busia, Budalangi, Funyula and Mumias were selected to reflect representative number of patients in the cohort. This region has an adult HIV prevalence of 6.7% (NASCOP 2007).

The study covered a cohort of 1000 patients on antiretroviral therapy supported by the Medecins Sans Frontieres (MSF) programs in Busia district hospital between 2007 and 2009. From this cohort, 75 persons were randomly selected. The samples analysed come from persons resident in the rural villages (Table 1) who, since they tend to travel less often, were very likely to have acquired their HIV infection in their villages. This facility serves patients from both Kenya and Uganda. Inclusion criteria comprised having been enrolled on antiretroviral therapy for at least 6 months.

After obtaining informed consent, venous blood collection was performed according to the protocol previously approved by the Kenya Medical Research Institute Scientific Steering Committee and Ethical Review Board (Ref. KEMRI SSC No. 1127). At least 5ml of whole blood were collected into vacutainer tubes from each study subject. The blood was packed and transported to the HIV lab in Nairobi for further analysis.

Peripheral blood mononuclear cells (PBMCs) were extracted from whole blood by density gradient centrifugation using Ficoll-Paque Plus (Pharmacia) and DNA extracted using DNAzol (Gibco BRL^®) and ethanol precipitation. The extracted proviral DNA was used for polymerase chain reaction (PCR) amplification. A region of the HIV-1 protease gene was amplified by nested PCR. Primers NYUPOL7 (5’-GGGAATTTTCTTCAGAGCAG-3’), and NYUPOL8 (5’-TCTTCTGTCAATGGCCATTGT-3’) were used in the first round and primers NYUPOL9 (5’-TCCTTAACTTCCCTCAAATCACT-3’) and NYUPOL10 (5’-CTGGCACGGTTTCAATAGGACT-3’)

were used in the second round to amplify 297 bp of the protease gene region corresponding to positions 2513 - 3209 in HIV-1 HXB2 sequence. The first round PCR was carried out in 25.0 ml tube containing 2.0 μl of DNA, 2.0 μl of 10× buffer, 0.4 μl of each forward and reverse primers (NYUPOL7 and NYUPOL8), 2.8 μl MgCl₂, 10.2 μl of distilled water and 0.2 μl of Taq polymerase. The amplification was carried out in a thermal cycler (MJ Research, Inc). Amplification was carried out with 1 cycle of 95˚C for 10 min and 35 cycles of 95˚C for 30 s, 30˚C for 60 s, and 72˚C for 1 min, and final extension of 72˚C for 10 min. The nested PCR used 2.0 μl of the first round PCR products as a template and used inner primers (NYUPOL9 and NYUPOL10) with the same cycling conditions. The PCR amplification was confirmed by ethidium bromide staining of samples electrophoresed on a 1.5% agarose gel.

The primers NYUPOL9 were used for PCR direct sequencing in an automated DNA sequencer (ABI 3100 Applied Biosystems, Foster City, CA) with BigDye Terminator version 3.0 Cycle Sequencing Reaction Kits (Applied Biosystems, Foster City, CA). The sequences were then aligned with previously reported HIV-1 strains of various subtypes from the Los Alamos HIV-1 database (http://hiv-web.lanl.gov) using CLUSTAL W profile alignment option. After sequencing the 75 samples, the generated nucleotides were analysed by BLAST to determine the HIV subtypes.

Genetic distances were calculated by the two-parameter method of Kumar et al., 2001 [16] . On this basis, only 64 samples were used to construct the phylogenetic tree by neighbour-joining method since the other sequences were too short (tree not shown). The reliability of the tree was estimated by 1000 bootstraps [16] [17] . The tree profile was visualized with Tree View PPC version 1.6.5. The generated sequences were deposited in

The table shows that most of the predominant HIV-1 isolates in this border area are subytpe A1, associated with Uganda.

the Genbank with accession numbers HQ176975-HQ177040.