Dynamics of cholera epidemics from Benin to Mauritania

Background The countries of West Africa are largely portrayed as cholera endemic, although the dynamics of outbreaks in this region of Africa remain largely unclear. Methodology/Principal findings To understand the dynamics of cholera in a major portion of West Africa, we analyzed cholera epidemics from 2009 to 2015 from Benin to Mauritania. We conducted a series of field visits as well as multilocus variable tandem repeat analysis and whole-genome sequencing analysis of V. cholerae isolates throughout the study region. During this period, Ghana accounted for 52% of the reported cases in the entire study region (coastal countries from Benin to Mauritania). From 2009 to 2015, we found that one major wave of cholera outbreaks spread from Accra in 2011 northwestward to Sierra Leone and Guinea in 2012. Molecular epidemiology analysis confirmed that the 2011 Ghanaian isolates were related to those that seeded the 2012 epidemics in Guinea and Sierra Leone. Interestingly, we found that many countries deemed “cholera endemic” actually suffered very few outbreaks, with multi-year lulls. Conclusions/Significance This study provides the first cohesive vision of the dynamics of cholera epidemics in a major portion of West Africa. This epidemiological overview shows that from 2009 to 2015, at least 54% of reported cases concerned populations living in the three urban areas of Accra, Freetown, and Conakry. These findings may serve as a guide to better target cholera prevention and control efforts in the identified cholera hotspots in West Africa.


Methodology/Principal findings
To understand the dynamics of cholera in a major portion of West Africa, we analyzed cholera epidemics from 2009 to 2015 from Benin to Mauritania. We conducted a series of field visits as well as multilocus variable tandem repeat analysis and whole-genome sequencing analysis of V. cholerae isolates throughout the study region. During this period, Ghana accounted for 52% of the reported cases in the entire study region (coastal countries from Benin to Mauritania). From 2009 to 2015, we found that one major wave of cholera outbreaks spread from Accra in 2011 northwestward to Sierra Leone and Guinea in 2012. Molecular epidemiology analysis confirmed that the 2011 Ghanaian isolates were related to those that seeded the 2012 epidemics in Guinea and Sierra Leone. Interestingly, we found that many countries deemed "cholera endemic" actually suffered very few outbreaks, with multi-year lulls. PLOS  Introduction Seven cholera pandemics have been documented since 1817 [1]. The disease has plagued every continent, spreading along trade routes via both land and sea [1]. Current epidemics are however localized to South Asia, Haiti, and Sub-Saharan Africa [2]. Since the ongoing pandemic first reached West Africa in 1970 [1], outbreaks have been repeatedly reported throughout the region [2]. However, only a few small-scale studies have investigated the dynamics of recent cholera epidemics in West Africa. Overall, cholera outbreaks throughout the study region have displayed markedly diverse patterns depending on the country. Certain countries such as Benin and Togo have reported cholera cases nearly every year albeit with relatively low incidence [3], [4]. By contrast, many other countries to the northwest such as Gambia, Senegal, and Mauritania have experienced marked multi-year lulls [5][6][7]. Many large epidemics erupted on the heels of violent civil conflicts that engendered humanitarian and public health crisis or massive population movement such as that in Senegal in 2004-2006 [7]. The large majority of cases were also reported from large cities following increased rainfall [7][8][9][10]. Although many studies were limited to a single outbreak/neighborhood or a short time period, common risk factors were found across the region: crowded living conditions, poor sanitation, and limited access to potable drinking water [3], [8][9][10][11]. Cholera is contracted by consuming food or water contaminated with toxigenic Vibrio cholerae O1 or the derivative Vibrio cholerae O139 [12]. Numerous V. cholerae non-O1 and some O1 serogroups lacking the cholera toxin are autochthonous in seawaters worldwide [13]. V. cholerae non-O1 serogroups have been found associated with a variety of aquatic flora and fauna, notably copepods [14]. In the Bay of Bengal, elevated seawater temperatures, copepod and plankton blooms, and rainfall have been shown to correlate with increased concentrations of V. cholerae in the environment [15][16][17]. In West Africa, a study has addressed the relationship between climate inter-annual variability and cholera in Nigeria, Benin, Togo, Ghana, and Ivory Coast over a 20-year period. From 1987-1994, they observed temporospatial synchrony between cholera incidence and rainfall in all countries except Ivory Coast [18]. Cholera has thus been depicted as a waterborne disease driven by ecological factors [12], [19]. However, despite technological improvements, a perennial aquatic reservoir of cholera-causing V. cholerae O1 has yet to be identified in West Africa [20].
We applied an integrated approach to describe the dynamics of cholera epidemics, including the identification of hotspots, and investigate factors that may influence the disease in several coastal West African countries from Benin to Mauritania. We analyzed weekly cholera outbreak evolution (at the district or commune level) throughout the region between 2009 and 2015 and field investigations in Benin, Togo, Ghana, Ivory Coast, Sierra Leone, and Guinea. We performed molecular epidemiology analysis (MLVA (Multi-Locus VNTR [Variable Number Tandem Repeat] Analysis) and whole-genome sequencing) of V. cholerae isolates from the majority of outbreaks affecting the study region since 2010. Such molecular epidemiology analysis can supplement epidemiological findings to provide further insight into the relationship between V. cholerae isolates and epidemic populations, identify clusters, establish phylogeny, and track bacterial transmission. We describe our findings country by country, from Benin to Mauritania.

Cholera case and rainfall data
Databases of all suspected cholera cases were collected from the epidemiological units of Benin, Togo, Ghana, and Guinea. National databases comprised weekly case/death numbers at the district level (Ghana and Togo), commune level (Benin), or prefecture level (Guinea) since 2009, according to the WHO cholera case definition (S1 Text). For Ghana (Accra), Togo (Lomé), Benin (Atlantique and Littoral/Cotonou), and Guinea (Conakry), we also analyzed the cholera case line lists, which include data on age, sex, clinical outcome, and residence. We obtained approval from the Ministry of Health (MoH) of each country to use these databases for epidemiological, research, and publication purposes. The line lists were anonymized and cleaned prior to analysis.

Field visits
For each field visit carried out in Ivory Coast, Ghana, Togo, and Benin, we first organized and coordinated the study with UNICEF and WHO (for Ivory Coast), through whom we were put in contact with national authorities to obtain official access to databases and V. cholerae strains (when possible). Initial field visits involved contact with these local UNICEF and/or WHO offices as well as national public health (surveillance and laboratory) authorities. To identify cholera hotspots and sites that may play a role in cholera diffusion, we first analyzed outbreak histograms and mapped weekly cholera cases. Field visits were then performed in the identified locales (e.g., sites that were repeatedly affected by cholera outbreaks or sites of an initial outbreak). During field visits, and accompanied by local counterparts, we met with local surveillance departments, laboratories (for information concerning lab-based case confirmation), and health facilities. Information was collected concerning how cholera was contracted as well as possible links with other cases. We also evaluated local WASH (Water, Sanitation, and Hygiene) conditions. Field visits were performed in Ivory Coast (12/2013; SM and RP); Ghana, Togo and Benin (11-12/2014; SM, PC, and RP); Guinea and Sierra Leone (08-09/2012; SR). See S1 Text for additional details concerning the field visit study protocol.

Ethics
The study was approved by the MoH of each country where field visits were carried out. The protocol was further approved by the Ghana Health Service Ethical Review Board according to standard procedures. The remaining countries did not seek ethics approval as epidemic disease surveillance and response is covered by national public health laws as an integral part of the public health mandate of each MoH.
All clinical isolates studied were analyzed anonymously.

Cartography
The country maps were generated using QGIS v2Á8-Wien with shapefiles from DIVA-GIS (http://www.diva-gis.org/gdata). The shapefile of Accra was generated with QGIS in collaboration with the Ministry of Local Government, Division of Environmental Health, Accra.

Vibrio cholerae isolate culture, DNA isolation, MLVA, and whole-genome sequencing
To design effective public health strategies to combat cholera, it is critical to understand the mechanisms of cholera emergence and diffusion in a region-specific manner. Genetic analysis of responsible strains can supplement epidemiological findings to provide further insight into the relationship between pathogenic strains and epidemic populations [21]. Indeed, isolate genotyping is useful to differentiate between different isolates, identify clusters, establish phylogeny, and track bacterial transmission. Lam et al. [22] have shown that MLVA represents a highly discriminatory technique to distinguish between closely related seventh pandemic isolates. They have also emphasized that the method is best applied for outbreak investigations or to identify the source of an outbreak. Rebaudet et al. have recently demonstrated that MLVAbased analysis of clinical V. cholerae isolates combined with an epidemiological assessment was instrumental in deciphering the origin of the 2012 cholera epidemic in Guinea [10]. All V. cholerae isolates were selected in a manner as to represent both the spatial and temporal evolution of each cholera epidemic in each country sampled. A total of 173 V. cholerae O1 clinical isolates collected throughout Ghana from 2010 to 2014 were provided by the National Public Health Reference Laboratory, Accra. The isolates were sub-cultured and transported in glycerol tubes at ambient temperature to Marseille, France. Aliquots of the culture were directly submitted for DNA extraction. We also analyzed three V. cholerae isolates from Senegal in 2011 following the same procedure. The MLVA results of the Ghanaian isolates were compared with those of previously analyzed strains from Togo (35 isolates), Guinea (37 isolates), and Sierra Leone (9 isolates) as previously described [10], [23].
DNA was extracted using a NucliSENS easyMAG platform (bioMérieux). MLVA of the isolates using an ABI PRISM 3130 Genetic Analyzer (Applied Biosystems) was performed using six VNTRs as described previously [23], and the relationship between the isolates was established using the goeBURST algorithm on PHYLOViZ v1.1 (http://www.phyloviz.net/). We performed a phylogenetic assessment of the core V. cholerae genome of the strains of the third wave of the seventh pandemic [24] based on genome-wide SNPs. Isolates from Togo-2010 (six), Togo-2011 (six), Togo-2012 (five), Ghana-2011 (one), Ghana-2014 (five), and Guinea-2012 (two) were also included in the analysis. DNA was sequenced using a HiSeq Illumina System (Illumina) and analyzed as described [24].

Cholera epidemics from Benin to Mauritania
From 2009 to 2015, Benin and Togo accounted for a combined average of 694 reported cholera cases annually (Table 1). In Benin, the lakeside commune of Sô-Ava, which is directly connected to Nigeria via Lake Nokoué and Yewa River, reported cholera outbreaks every year since 2010 and was often the first and hardest-hit commune. In   Once outbreaks erupted in Accra, cholera rapidly diffused throughout the majority of the city. This rapid spatial diffusion pattern was observed during the onset of the 2011, 2012, and 2014 epidemics. Many Accra neighborhoods were severely affected by cholera each year. However, certain nearby residential areas remained largely cholera-free, despite outbreaks in adjacent neighborhoods (Fig 2). Once cholera erupted in Accra, outbreaks spread to other districts in Ghana several weeks later. According to health facility staff at the hospital in Ho (Volta Field visits revealed that water distribution was often interrupted for several days in many Accra neighborhoods. The majority of water network pipes were visibly damaged and running along the ground through roadside gutters. Residents without access to proper latrine facilities perform open defecation into these gutters. In the Greater Accra Metro Area, access to improved sanitation facilities is limited. Only 5.7% of households in Greater Accra have access to latrines that flush into a piped sewer system, while 35.1% of Accra households use latrines that flush into a septic tank. Furthermore, Accra households lacking improved sanitation facilities (14.8%) are forced to use pan latrines (buckets that are then dumped into the roadside gutters) or "flying toilets" (defecation into a plastic bag which is then literally thrown away) [25]. We thus hypothesize that ground water and human waste could seep into broken pipes, especially during the frequent water shortages, thereby allowing V. cholerae to enter the water network and spread when water pressure is restored. Increased rainfall markedly exacerbated this effect (Fig 1 and Fig 2). Additional studies should further investigate the role that the Accra water network plays in cholera outbreaks to confirm this hypothesis. Following a three-year lull in the incidence of cholera, both Sierra Leone and Guinea experienced a trans-border epidemic in 2012, with 22,932 and 7,351 cases, respectively [10]. The two epidemics progressed following a very similar pattern. In Guinea, the outbreak started in February on Kaback Island with a fisherman traveling from Sierra Leone. In Sierra Leone, possible events of V. cholerae importation by fishermen travelling from Liberia and Ghana have been reported [29]. During the rainy season, cholera exploded in the capitals, which recorded over half of the total cases (Freetown, 52%; Conakry, 64%) [10], [29].
As cholera rates declined in Sierra Leone and Guinea, they rose in Guinea-Bissau [29][30][31], which also followed a near three-year lull. The country reported 3,068 cases in 2012 and 969 cases in 2013. Eighteen and zero cases were reported in 2014 and 2015, respectively [2].
The number of cholera cases reported in The Gambia, Senegal, and Mauritania has been very low since 2009 to present. The Gambia has not reported a single suspected cholera case since 2008. Likewise, Mauritania has not reported cholera cases since 2008, with the exception of 46 cases in 2011. Since 2009, Senegal has reported only 13 suspected cholera cases [2].

Genetic analyses of V. cholerae isolates
We performed MLVA of 255 clinical V. cholerae isolates from Ghana, Togo, Guinea, Sierra Leone, and Senegal. Two environmental isolates from Guinea 2012 were also included (S1 Table). Interestingly, the Minimum Spanning Tree (MST) shows that the 2010/2011 isolates from Ghana were related to those that seeded the epidemic in Guinea and Sierra Leone in 2012. The MST also demonstrates that the 2011, 2012, and 2014 epidemics in Ghana were due to three distinct V. cholerae MLVA-type clusters. The three clinical isolates from Senegal in 2011 displayed an identical MLVA type, closely related to strains from Togo in 2011 and 2012, which may represent imported cases from farther south in West Africa (Fig 3).
When the Ghana and Togo strains were included in the alignment of the seventh pandemic V. cholerae isolates [24] (Fig 4).

Discussion
From 2009 to 2015, we found that Accra, the capital of Ghana, reported the highest number of cholera cases during the study period among all cities included in this study. During the period 2009 to 2015, we found that one major wave of cholera outbreaks spread from Accra in 2011 northwestward to Sierra Leone, Guinea, and likely Guinea-Bissau in 2012 (S5 Fig). Genetic analysis showed that the 2012 isolates from Guinea and Sierra Leone clustered with those collected in Ghana in 2011. MLVA also demonstrated that the V. cholerae strain responsible for the epidemic in Ghana, which started in 2010 and spilled over into 2011, was already present in Togo and thus likely shared a common ancestor with strains from Togo in 2010. The MST and whole-genome sequence results indicate the presence of different V. cholerae populations in Togo, one group that appears to give rise to the 2011 Ghana strains, while the other group was closely related to the strain that triggered the cholera epidemic in Ghana in 2014. We noted that other cities (Conakry and Freetown) also appeared to function as amplifiers of cholera, when cases were present and rainfall increased. Strikingly, we found that many countries deemed cholera endemic in modeling studies [32] actually suffered very few outbreaks, with  multi-year lull periods during which no cholera cases were detected. Extended lulls in cholera incidence occurred despite increased rainfall, typical high temperatures, slums, and population exposure to coastal environments. According to the WHO, cholera endemic countries have been defined as countries in which confirmed cholera cases were reported in at least three of the five past years [32]. In fact, some of the countries in our study do not fit the definition of endemic, as at least one case must be confirmed. To improve the current classification methodology, we may consider repeated, lab-confirmed cholera outbreaks in a country to indicate endemicity, keeping in mind that this status is fluid and may change as sanitation and water conditions improve.
Our findings and independent reports indicate that the Accra water network may play a role in rapid diffusion of cholera throughout a majority of the city, when cholera cases are present in neighborhoods where sanitary facilities and access to safe water are lacking. A study in Accra (in Osu Klottey Sub Metro) has shown that drinking community pipe-borne water (OR = 2.15) was associated with cholera in 2012 [33]. Furthermore, a separate study has revealed unsuitable residual chlorine levels and the regular presence of fecal coliform in the Accra network water [34]. During the period 2009 to 2015, our findings show that one major wave of cholera epidemics spread northwestward from Accra in 2011 to Sierra Leone and Guinea in 2012. As Ghanaian strains from previous years fail to reappear during subsequent epidemics, we hypothesize that epidemics affecting Accra may likely originate due to imported cases from a nearby cholera hotspot. Neighboring Nigeria represents one of the major cholera foci in the world [35], with 130,007 cases reported from 2009 to 2015 [2]. The country is also a likely source of outbreaks in the Lake Chad Basin [35]. The lull in Ghana during 2013 paralleled a relatively low number of cholera cases reported in Nigeria in 2012 and 2013. The 2014 Ghanaian epidemic coincided with an epidemic rebound in Nigeria [2]. Furthermore, intense commercial activity via road and boat may represent a major pathway by which cholera is imported from outbreaks in Nigeria to susceptible waterfront communities in Benin. We hypothesize that the detection of closely related strains in Ghana and Togo throughout the study period indicates that this strain may share a common ancestor with strains responsible for persisting outbreaks in a neighboring cholera hotspot such as Nigeria.
A phylogenetic analysis of clinical isolates has shown that the current pandemic is characterized by successive global clonal expansion of three waves of closely related serotype O1 El Tor lineages emanating from the Bay of Bengal [24]. A recent study in Science by Weill et al. has analyzed genomic data from 1070 V. cholerae O1 isolates, across 45 African countries and over a 49-year period, to show that past epidemics were attributable to a single expanded lineage. Weill et al. show that seventh cholera pandemic V. cholerae El Tor sub-lineages from Asia were repeatedly introduced into West Africa as well as East and Southern Africa. Epidemic waves then propagated regionally over a period of several years, which correlates with our observations [36]. Their findings strongly suggest that human factors play a much more important role in cholera dynamics and the long-term spread and maintenance of V. cholerae in Africa than environmental factors [36]. Further whole-genome sequence analysis of strains circulating in West Africa would greatly enhance our understanding of V. cholerae transmission pathways in the region.
Concerning the study limitations, although the databases used in this study are reliable thanks to cholera surveillance standardization throughout the region (S1 Text), the system does not guarantee detection of every single sporadic or unique imported cholera case. However, cholera cases are quickly detected once an outbreak occurs, especially when cholerarelated deaths result. Many countries deemed cholera endemic have experienced several yearlong lulls in cholera outbreaks (when all suspected cholera cases are confirmed negative upon culture). Nevertheless, countries considered cholera endemic anticipate outbreaks and perform culture-based tests on samples derived from suspected cholera cases each year. Given the capacity of the countries concerned (which can vary over time and geographically within a country), it is not possible to bacteriologically confirm every suspected case, which could have an effect on the validity of the results. Similar issues concerning surveillance capacity may have affected the completeness of suspected cholera case data collection over time. Nevertheless, our investigations have found that the lack of cholera during a lull period is not due to inadequate surveillance or laboratory incompetence but simply due to the absence of cholera cases in the country. The nonexistence of cholera outbreaks during the recent Ebola crisis in Guinea, Sierra Leone, and Liberia highlights the complete lull in cholera cases despite a heightened disease surveillance system. Furthermore, all isolates analyzed via MLVA were not submitted for whole-genome sequencing due to inadequate shipping conditions, which yielded sufficient DNA for only PCR-based MLVA. V. cholerae isolates from countries such as Ivory Coast, Benin, Guinea-Bissau, and Liberia were not included as the concerned national laboratories did not provide isolates for this investigation. We also acknowledge that our study would be strengthened by sequence-based phylogenic results of all samples assessed via MLVA, and thus efforts are currently underway to complete our panel of V. cholerae isolates. Furthermore, we were not able to include all isolates for sequence analysis, as certain samples lacked sufficient concentrations of DNA for the technique, thus introducing a possible bias due to incomplete selection of samples for whole-genome sequencing.
To prevent expansion of cholera outbreaks in the analyzed region of West Africa, epidemiological surveillance should be enhanced in identified vulnerable zones, such as Accra. In vulnerable areas, improved monitoring of the drinking water supply as well as ensuring water quality and proper chlorination would mitigate epidemics and perhaps stop cholera propagation.
Based on our findings, we hypothesize that once cases arrive in the urban settings with poor sanitation facilities (as observed in Accra, Conakry, and Freetown), increased rainfall provokes the infiltration of human waste, and therefore toxigenic V. cholerae, into the water network via damaged water pipes, thus promoting a rapid increase in cholera incidence. These findings may serve as a guide to better target cholera prevention and control interventions in the identified cholera hotspots in West Africa. Our study also highlights the value of this type of study combining epidemiological and molecular data to gain insight into the dynamics of cholera, especially in Africa.  Table. The epidemic populations, isolate IDs, and PCR amplicon size of each allele corresponding to each MLVA type. The number of isolates corresponding to each MLVA type is indicated on the right. The environmental isolates are indicated with an asterisk (MLVA type column). (DOCX)