Each of the 76 brain samples used in this study was positive for RABV antigen. The overall topology of the phylogenetic tree produced by our analysis of the RABV N-gene sequence data available from a sample of rabid African dogs and cats in Ghana was consistent with those previously described 
. This analysis of Ghanaian rabies cases is the first phylogenetic analysis of RABV from Ghana. Where this analysis is distinct from reports of RABV in other West African nations is in the diversity of viruses detected within Ghana. The samples were all taken from a relatively small geographical region with those samples not from within the greater Accra region originating from towns relatively close to Accra. These included eight viruses from Tema and five from Cape Coast (25 and 142 km from Accra, respectively). There was no evidence of infection with Africa 3 RABV (detected in mongoose in southern Africa) 
, Africa 4 RABV (detected in north-eastern Africa) 
or other Lyssavirus
species such as Lagos bat virus, against which a high seroprevalence of antibodies has been detected in bats from Accra 
. However, our analysis suggests that rabies epidemiology is much more complex than at first thought from previous studies within West Africa. Indeed, whilst West African countries typically have defined lineages circulating within them, only Nigeria and the Central African Republic have previously been described as having Africa 1 and 2 lineage viruses co-circulating within their national borders 
. We detected both in Ghana, and propose that Ghana's recent history and geography may explain why both virus lineages were detected.
Africa 2 viruses appear to have been present within the dog populations of West Africa, including Ghana, for decades. This is derived from the close relationships between the RABV characterized in Ghana and those reported in other West African countries, such as Benin, Ivory Coast, Burkina Faso and Niger. Our results support the findings of others that the Africa 2 virus lineage has been circulating within Africa for less than 200 years 
. Within Ghana, our analysis suggests the Africa 2 clades now co-circulating in Ghana have different evolutionary histories. From the Africa 2 phylogenetic analysis (), we hypothesize that the three Ghanaian Africa 2 clades co-circulate in Ghana, but share evolutionary histories with viruses from other West African countries. Whilst we cannot be certain of the direction of the virus spread, we believe that there have been three different introductions of Africa 2 viruses to Ghana. We found support for the hypothesis that one clade that circulates in Ghana and in the northeasterly West African countries of Niger and Burkina Faso was originally imported from Niger and subsequently entered both Ghana and Burkina Faso (). Another clade of viruses share a common ancestry with a Beninese isolate from the east and likely entered the country from Benin or via neighboring Togo. The evolutionary history of those viruses from the east and northeast may be due to Lake Volta providing a physical obstacle to virus transmission between dog populations. Further analysis of this phylogenetic relationship is precluded, however, by the lack of additional published sequences from Benin, and none from neighboring Togo. A single virus, G6, forms a clade with isolates from the Ivory Coast. This virus appears to be a recent introduction, sharing a TMRCA of just 1 to 20 years with viruses from the Ivory Coast to the west. A possible reason for fewer viruses being from the Ivory Coast may be the large tropical forest system along the Ghana-Ivory Coast border providing a barrier to dog movements. The border with the Ivory Coast was historically the most forested area of Ghana, however rapid deforestation and increasingly easy “between country” travel may have led to the trans-boundary movements of this virus.
Due to the historical dominance of Africa 1 viruses in the northern, eastern and southern parts of Africa, we believe it reasonable to hypothesize that Africa 1 viruses have entered Ghana from those regions, and that transmission has not been from Ghana to those regions. This hypothesis is supported by the phylogeographic analysis which suggests that the virus sub-lineage Africa 1a was transmitted from central African counties to Ghana. If we accept this, the origin of the Ghanaian Africa 1a sub-lineage viruses may be explained simply by virus transmission through dog (and potentially other vector) populations from central African nations to Ghana (). Indeed, in our analysis the Ghanaian Africa 1a viruses share an ancestry with a virus from Gabon with a TMRCA estimated to be 23–31 years ago. This would require viruses to be transmitted at an approximate rate of between 39 to 53 kilometers per year. The large number of Africa 1a viruses in our sample suggests that this sub-lineage is well established in the Accra region, however further virus sequences from nations between Ghana and Gabon are required to confirm the evolutionary history of this sub-lineage.
The presence of an Africa 1b sub-lineage RABV in our analysis is the first reported from West Africa. Analysis of the Africa 1 lineage viruses suggests that this virus shares an ancestry with viruses from East Africa, in particular, those from Kenya (). The presence of this virus may be explained in one of two, not exclusive, ways. Firstly, sub-lineage 1b viruses may simply have been transmitted within the populations of dogs and other susceptible animals from eastern African countries to Ghana. Transmission from Kenya (with Nairobi approximately 4200 km from Accra) would require virus transmission at a rate of approximately 190–279 kilometers per year with the TMRCA estimated to be 18 years (15–22 years, 95% HPD). Given the distance infected dogs and potential wildlife hosts may travel, this is theoretically possible, but highly unlikely given that rabies spread in red foxes and raccoons in Europe and North America was estimated to be typically 30–60 kilometers a year 
. Therefore, we hypothesize that the more likely reason for this virus' presence in Ghana is that an infected animal was translocated from the east, thus introducing a new sub-lineage to the region. Indeed, we believe that this may be the first report of molecular evidence of a long distance translocation of a rabies sub-lineage in Africa.
Spatio-temporal models of rabies in eastern and southern Africa show large-scale synchrony of rabies epidemics across both regions 
. The analysis by Hampson et al
provided evidence that movement of infectious animals, or animals in the incubation period, and localized regional or national vaccination campaigns during epidemics, are likely to lead to rabies synchrony 
. However, evidence provided by rabies control programmes in both Europe and the Americas show that large-scale control programmes can be successful 
. A study of rabies in Tanzania also suggested dog rabies control was feasible, but was hampered by perceived problems that were largely unfounded 
. A subsequent analysis by Hampson et al
suggested that regular regional pulsed vaccination programmes would be required to eliminate dog rabies 
. Despite the analysis estimating the basic reproductive rate of domestic dog rabies throughout the world to be low (R0<2), the rapid turnover of dog populations led to enough susceptible hosts for rabies to be maintained 
. Our molecular study suggests introductions of RABV from neighboring countries into Ghana are not infrequent, demonstrating that without substantial support for continuous vaccination or coherent regional cooperation, Ghana will be unable to eliminate rabies and maintain a rabies-free status. In addition to this, our analysis provides evidence of a virus that shares a recent common ancestry with viruses from East Africa, therefore providing further evidence that regional control programmes must be implemented and that once rabies is eliminated, vigilance and technical expertise must be maintained in order for new introductions to be controlled 
Currently rabies diagnostics within the Ghanaian veterinary services remain limited to non-Lyssavirus
species specific staining techniques, including the Sellers' stain and, when FITC conjugate is available, FAT. Inadequate government and financial commitments and a resource limited veterinary infrastructure are restrictive factors that preclude a sustainable rabies diagnostic service in Ghana. Surveillance activities should be given a higher priority to maintain an effective diagnostic service with the co-operation of other national and international organizations. Each of the 76 brain samples used in this study was positive for RABV infection by RT-PCR at VLA Weybridge. Of the 76 samples full histories were available for 72 positive rabies cases. However, seven samples were negative when tested by Sellers' stain at the VSL. The VSL recorded 66 humans being bitten by those 72 dogs for which histories were recorded (data not shown), including six bites to humans by the seven RABV positive cases that tested negative in the VSL. Further training and the availability of FITC conjugate for the FAT or use of the direct rapid immunohistochemical test (dRIT) 
may have overcome some of the diagnostic problems. However, given that low cost isothermal RT-LAMP assays have been developed for a number of viruses affecting livestock in Africa, including Rift Valley Fever virus 
and African Swine Fever virus 
, we developed and tested the RT-LAMP for use in African laboratories. The RT-LAMP may be prone to some of the same problems as other molecular techniques, such as cross-contamination, however it is a cheap molecular technique that produces a product that is available for further analysis such as sequencing of the approximately 200 bp product. We developed the novel RT-LAMP on randomly selected RABV samples, including both Africa 1 (a cosmopolitan) and 2 lineages. This assay successfully amplified viral genetic material producing a measurable DNA product for both Africa 1 and 2 lineage viruses. This isothermal diagnostic assay negates the need for thermal-cyclers for molecular diagnosis of RABV. The assay reagents costs approximately $3 per assay and therefore may prove a useful alternative assay for those laboratories that already have molecular expertise and adds to the range of rapid cost-effective diagnostic assays that will be fundamental if developing countries wish to develop their own RABV diagnostic capabilities. Whilst “snap test” LFD tests have previously been reported 
our adaptation of the RT-LAMP assay to use an LFD platform, instead of UV illumination, further reduces the technology required for RABV diagnosis in African laboratories. Additional validation of this method will require comparison with the gold standard assays, assessment of larger panels of samples from throughout Africa, as well as evaluation of its sensitivity in detecting RABV in brain samples from OIE reference laboratories. These preliminary findings, however, demonstrate proof-of-concept and suggest that this technique has the potential to provide African laboratories with a cheap and rapid molecular detection method.
We conclude that our analysis of rabies virus sequences derived from Ghana has furthered the understanding of RABV epidemiology in West Africa. In particular, our analyses suggest that both Africa 1 and Africa 2 RABV lineages are present in Ghana. Africa 1b sub-lineage had previously not been reported in West Africa, and its detection, along with evidence of an additional four further clades circulating in Ghana support previous analyses that suggest that only sustained regional level approaches to rabies control will be successful in rabies elimination. In addition, we have developed an African RABV RT-LAMP assay, which can be adapted for use with LFD platforms that we advocate will provide an additional diagnostic tool for African regional laboratories.