Category Archives: epidemiology

Microbial Forensics of a Natural Pneumonic Plague Outbreak

For bioterrorism agents like Yersinia pestis it is necessary to identify the strain and its source specifically enough for forensic use. Categorizing an epidemic isolate and tracing its source is always important for public health measures, but the level of precision is far higher for legal uses. Developing forensic techniques to characterize and parse very similar strains of a species and trace it to a specific location robs terrorists (and states) of the ability to deny responsibility for an attack (Koblentz & Tucker, 2010). The ability to launch a secret and deniable attack on an enemy has been viewed as one of attractive advantages of biological warfare.

A Chinese group led by Ruifu Yang and Yujun Cui recognized that only whole genome sequencing could adequately parse the strains of the monomorphic species Yersinia pestis but that the computing power necessary to compare entire genomic sequences as the database enlarges is impractical (Yan et al, 2014). Unlike most pathogens, typing only specific regions of the genome are just not enough to get a unique genetic fingerprint for low genetic variability pathogens like Yersinia pestis. This is yet another indication of the genomic similarity of all Yersinia pestis strains.

The Chinese group developed a two stage method of classification detailed enough for forensic work.  They took a twelve person outbreak of pneumonic plague contracted from a dog in 2009 in the Qinghai area of Tibet / western China, specifically at Xinghai as their test case (Wang et al, 2010). In the first step they took six cases including the two dogs who died in the outbreak and compared them to 24 strains representing the 23 phylogroups of the phylogenetic tree. This comparison selected which branch of the phylogenetic tree the outbreak belonged. There were no SNP (single nucleotide polymorphisms) different between the seven isolates confirming a common source, one of the dogs based on outbreak narratives. The seven isolates were all the same strain belonging to branch 1.IN2 of the tree. The second step was to then compare the isolates to all known strains of 1.IN2 shown below. Since these strains all come from the Qinghai-Tibetan plateau, they were able to add other strains historically isolated from this region.

Distribution of 1.IN in Qinghai  (site source)
Distribution of 1.IN2 in Qinghai (Yan et al, 2014, click to enlarge)

The results localized the new isolates (r) as being from the same focus as strains g, r, s, t. u plus, interestingly, the 0.PE7 strain (green b) that is over 300 SNPs different from the 1.IN2 strains. All of these other strains from this branch are scattered around the Qinghai region near Lake Qinghai. The polysomy (branch point) that produced all of the 1.IN2 in Xinghai (g,r,s,t,u) is located closer to the eastern end of Lake Qinghai, where the Chinese team hypothesizes this these strains began. The new outbreak isolates did not match any previous isolates from Xinghai which is testimony to the degree of movement of these strains around the region. Without the case narrative, they would not have been able to identify the specific foci at Xinghai, but would have got it to the region of east Qinghai lake. This illustrates how important sampling all of these foci are because a biological attack is likely to be far from its site of environmental isolation. Characterization of all laboratory strains, obviously, needs to happen as well for forensic tracing.

Reconstructing the historical epidemiology of this region will be an area of continuing research. The location of 0.PE7, the most genetically ancestral strain ever found — the closest the common ancestor of all Yersinia pestis, plus the likelihood that the ‘big bang’ epidemic (or epizootic), that produced the third pandemic, represented by node 12, was also in this region. (Each of the nodes represents a bang of evolutionary diversity, with all major branch points in the lineage probably representing large epidemics or epizootics.) The full diversity of strains in this region (unrelated to the outbreak isolates) are not shown in the figure above. This same group lead by Ruifu Yang  produced the primary phylogenetic tree of Yersinia pestis in China that noted that the molecular clock is not constant (Cui et al, 2012), here calculates that N12 is about 212 years old (95% confidence being 116 to 336 years ago) (Yan et al, 2014).  They note that in the history of Qinghai, there was a major human outbreak in the year 1754 CE linked to a Buddhist missionary working in Qinghai and Gansu provinces (Yan et al, 2014). Its is unclear if we can trust this narrative at all; scapegoats are common in plague narratives. Linking the 1.IN2 strains from Qinghai to four of the five o.IN2 isolates from Tibet suggest that the epidemic moved from Qinghai to Tibet in one ancient epidemic, though remaining isolate from Tibet looks like a more recent transmission from Qinghai. Regardless of the movements of 1.IN2, this area is believed to have been a site of long-term survival of Yersinia pestis, potentially over a thousand years, so that it has a lot to teach us about enduring foci.

Microbial forensics has already been used in criminal investigations, court cases and intelligence operations, such as the ‘Amerithrax’ (anthrax) attacks of 2001, anthrax spores sprayed over Japan by a cult, and suspicious plague cases in New York City (Yan et al, 2014). Phylogenetic microbial forensics was successfully used to show the intentional transmission of HIV from Dr Richard Schmidt to his girlfriend in his 1998 trial. This was the first successful use of microbial forensics in a court case (Koblentz & Tucker, 2010). In these cases, isolates are taken from the accused, the victim, other sexual partners, and the local population so show phylogenetic linkage between the accused and victim in the context of the local epidemiology.  The United States, United Kingdom, Sweden, the Netherlands, Japan, Canada, Germany, Australia, Singapore, and now China are involved in the development of microbial forensics (Koblentz & Tucker, 2010; Yan et al, 2014).

Reference

Koblentz, G. D., & Tucker, J. B. (2010). Tracing an Attack: The Promise and Pitfalls of Microbial Forensics. Survival, 52(1), 159–186. doi:10.1080/00396331003612521

Yan Y, Wang H, Li D, Yang X, Wang Z, et al. (2014) Two-Step Source Tracing Strategy of Yersinia pestis and Its Historical Epidemiology in a Specific Region. PLoS ONE 9(1): e85374. doi:10.1371/journal.pone.0085374

Wang, H., Cui, Y., Wang, Z., Wang, X., Guo, Z., Yan, Y., et al. (2010). A Dog-Associated Primary Pneumonic Plague in Qinghai Province, China. Clinical Infectious Diseases, 52(2), 185–190. doi:10.1093/cid/ciq107

Cui, Y., Yu, C., Yan, Y., Li, D., Li, Y., Jombart, T., et al. (2012). Historical variations in mutation rate in an epidemic pathogen, Yersinia pestis. Proceedings of the National Academy of Sciences, 110(2), 577–582. doi:10.1073/pnas.1205750110/-/DCSupplemental/sd01.xls

General Principles of Zoonotic Landscape Epidemiology

Zoonoses, pathogens with animal reservoirs, exist as part of a complex system of interactions between animal reservoirs, vectors, ecological factors and human interaction. Landscape epidemiology has existed as a field of study since Russian epidemiologist E.N. Pavlovsky coined the term and laid the groundwork in the 1960s. Landscape epidemiology is in essence the study of environmental foci of zoonotic disease, what Pavlovsky called a nidas. Many of the variables have been identified and studied in individual pathogen systems.

Each system seems so complex and unique that it can be easy to think that they each exist as separate entities with little to do with each other. It is necessary to develop some general principles to both see the bigger picture, and guide research and response to less studied and newly discovered pathogens. Lambin et al. set out to do just that by doing a meta-analysis of eight regional case studies of zoonotic diseases in Europe and East Africa: West Nile Virus in Senegal, Tick-borne Encephalitis in Latvia, Sandfly abundance (leishmaniasis vector) in the French Pyrenees, Rift Valley Fever in Senegal, West Nile Virus hosts in Camargue, Rodent-borne Puumala hantavirus in Belgium, human cases of Lyme borreliosis in Belgium, and risk of malaria re-emergence in Camargue. Obviously, as indicated, not all of these studies look at all factors involved in landscape epidemiology so validation is not solely based on the number of case studies that support each principle.

The ten proposed principles by Lambin et al are shown graphically below where they fit into the system of variables.

Graphical representation of the landscape determinants of disease transmission. The numbers refer to the ten propositions formulated in this paper. Lambin et al. International Journal of Health Geographics 2010 9:54   doi:10.1186/1476-072X-9-54
Graphical representation of the landscape determinants of disease transmission. The numbers refer to the ten propositions formulated in this paper.
Lambin et al. International Journal of Health Geographics 2010 9:54 doi:10.1186/1476-072X-9-54

Proposed general principles (Lambin et al, 2010):

  1. Landscape attributes may influence the level of transmission of an infection” This proposal is found in all case species. Features of the landscape influence vector and host distribution across the region of study. Distribution and type of water (fresh, brackish, or salt water) is a common landscape feature that influences density of insect vectors.
  2. Spatial variations in disease risk depend not only on the presence and area of critical habitats but also on their spatial configuration“.   The sheer size of the critical area is not the only or necessarily the most important characteristic to determine risk in an area. Some vectors like ticks thrive along border zones between ecosystems, like edges between woodland and grasslands.
  3. Disease risk depends on the connectivity of habitats for vectors and hosts” Creating contact zones or contiguous zones that create linked areas are also important. The spatial configuration can create corridors for disease persistence in harsh landscapes. Type and connectivity of  vegetation is as important as terrain for vector habitats. Connectivity between suitable habitat for rodents and insects allows the disease to spread from one patch to the next amplifying the pathogen to a level that increases risks of human transmission. Connections between patches of critical habitats allows for recolonization after local extinction.
  4. The landscape is a proxy for specific associations of reservoir hosts and vectors linked with the emergence of multi-host disease.” Their principle could be better fleshed out; their primary evidence coming from West Nile Virus (WNV). Like other multi-host pathogens, WNV has some hosts that are much more important than others for transmission across wide regions. In WNV migratory birds are a key to understanding its spread and epidemic dynamics. WNV is also an example of a disease with different proxies and amplification hosts in different regions of the world.
  5. To understand ecological factors influencing spatial variations of disease risk, one needs to take into account the pathways of pathogen transmission between vectors, hosts, and the physical environment.” Vector-borne diseases require direct contact between humans and the vector. For other zoonoses like hantavirus contact between humans and animal hosts can be via aerosols of material with rodent feces or dust containing rodent remains. For example, people have contracted hantavirus by vacuuming up rodent remains in homes. When estimating risk of transmission to humans, abiotic (non-living) environmental conditions that can preserve or transmit to humans have to be considered. Climate and moisture content of the soil are common abiotic factors to be concerned about. Additional support for this principle comes from the role of the rodent burrow system on plague (Yersinia pestis) hosts and vectors.
  6. The emergence and distribution of infection through time and space is controlled by different factors acting at multiple scales” In their discussion of this principle, they focus on human interaction with the environment and particularly urbanization altering disease risk. They note that climate change and natural environmental change do not account for all emerging and re-emerging disease but the activities of humans including urbanization and ecological change like deforestation. Ben-Ari et al‘s study on plague and climate change also looks at the many factors at all levels from micro to macro scales effect the abundance and likelihood of transmission of the plague.

    Plague cycle including hosts and vectors with abiotic influences
    Plague cycle including hosts and vectors with abiotic influences (Ben-Ari et al, 2011).
  7. Landscape and meteorological factors control not just the emergence but also the spatial concentration and spatial diffusion of infection risk” This principle just adjusts the previous principles to take account of primarily rainfall by looking at temporary ponds or wetlands. This particularly affects mosquito abundance, but as the graphic above demonstrates also effects soil moisture.
  8. Spatial variation in disease risk depends not only on land cover but also on land use, via the probability of contact between, on one hand, human hosts and, on the other hand, infectious vectors, animal hosts or their infected habitats” Land use has been long known to affect mosquito abundance and disease transmission. Clearing land for settlements or agriculture always increases standing water in ditches, tire ruts, railroad ditches, animal troughs, incomplete building projects, and due to loss of water absorbing vegetation. A century of malaria research and management has focused on land use and the elimination of standing water.  Mature water management programs for cultivation or flood control can also alter vector abundance and human contact rates. For example flooding fields to grow rice not only provides habitat for mosquito production but also brings people into the fields to cultivate increasing contact rates. Irrigation canals would have a similar effect.
  9. The relationship between land use and the probability of contact between vectors and animal hosts and human hosts is influenced by land ownership” In Lambin et al, they looked at the contact rates between public (state) land and private ownership. In these studies state ownership increased access to forestland over private ownership.By the same token, state ownership could also prevent deforestation and urbanization by preserving the wilderness or reserving the land for other uses. Forest age and maturity also varies significantly between state forests and private land.
  10. Human behaviour is a crucial controlling factor of vector-human contacts, and of infection.”  Humans bring themselves into contact with vectors by risky behavior and can control exposure vectors and infections. Obviously, vaccination is one of the controlling factors of infection, although many zoonotic infections have either no or poor vaccines. Occupational and recreational exposure to vectors often explains gender difference in infection rates.

In conclusion these principles begin to mark out the three sides of a zoonotic triangle: biology of pathogen, vector and host; ecological system where they exist; and human behavior and ecological interaction. Human behavior including land use and constructed environments is as important as the other two sides of the triangle. Humans are not passive victims or collateral damage.

Reference:

Lambin, E. F., Tran, A., Vanwambeke, S. O., Linard, C., & Soti, V. (2010). Pathogenic landscapes: Interactions between land, people, disease vectors, and their animal hosts. International Journal of Health Geographics, 9(1), 54. doi:10.1186/1476-072X-9-54 [open access]

Ben-Ari T, Neerinckx S, Gage KL, Kreppel K, Laudisoit A, et al. (2011) Plague and Climate: Scales Matter. PLoS Pathog 7(9): e1002160. doi:10.1371/journal.ppat.1002160

The Super-spreading Landscape of Urban Dengue Fever

Dengue Fever is one of the most concerning emerging infectious diseases of the early 21st century. The virus has been spreading with its ever-expanding host, the mosquito Aedes aegypti.  For the last several years there have been naturally acquired cases of dengue fever in the United States and Europe, that are not connected to travel.

Aedes aegypti from Tanzania (Source: Muhammad Mahdi Karim, 2009)

Aedes aegypti‘s preference for the urban environment distinguishes it from most mosquitoes. It prefers to lay it eggs in small urban pools of water – flower pots, old tires, car ruts, buckets – rather than natural forest pools. As day-light feeders, bed netting would not be useful against A. aegpyti.  It has been known for some time that A. aegypti populations are driven by super-producing sites, pools of water that produce the majority of mosquitoes vs. pools that only produce a very few pupae.

It is known that dengue fever is transmitted by super-spreading events but it is unclear how this is tied to A. aegytpi super-production sites and other factors in the environment. To study this phenomena a group of researchers from Yale School of Public Health and the Institiuto National de Salud in Bogota, Columbia chose a dengue fever endemic neighborhood to study the major parameters in transmission. They identified three primary parameters to monitor.

  1. Distribution of super-producing A. aegypti sites across urban plots with similar characteristics. (Mosquitoes generally remain very close to where they hatch and are believed to be exposed to the virus and transmit it within the plot they hatched, or at most a neighboring plot.)
  2. Density of humans domiciled in the plots who can be infected.
  3. Human to mosquito transmission of the Dengue Fever virus. (Humans are the primary reservoir.)

Padmanabha et al devised a new index they call the epidemic potential, secondary infection rate (Ro) per capita.   They hypothesize that human density alters the epidemic potential by altering the dengue viral introduction rate and the secondary infection rate.  Padmanabha et al. note that viral transfer from human to mosquito depends on the number of mosquito bites per person, while viral transmission to humans from mosquitoes depends on the number of different people an infected mosquito bites.

They selected 16 similar urban plots in an endemic neighborhood in Columbia with a range of 41 to 142 homes (1-3 city blocks) with a human density of 3.2 to 4.5 residents per house. They surveyed A. aegypti pupae in water containers to estimate mosquito production and trapped mosquitoes to look for infected adults. Humans immune response to dengue virus was also surveyed over the season. The mosquito surveys were conducted seven times and human immune surveys three times over the season.  They excluded schools, churches and other civil locations were the community gathers from the plots.

Mosquito density results demonstrated super-production sites in each of the seven surveys within each patch. Only 5% of the house surveys accounted for 92% of the total mosquito pupae found. Pupal abundance accounted for nearly 80% of the variation in vector production.  Their model predicted an Ro of 0.88 to 3.87 and correlated with the number of infected humans introductions that produced 20 or more secondary infections; this is only 10% of model repetitions. In most cases introduced viruses to the patch  did not produce secondary infections. Analysis of human-to-mosquito transmission (viral introduction to the patch) and mosquito-to-human transmissions (secondary infections) suggest that both human density and vector abundance alter the dengue Ro and epidemic potential. Models using data generated by this study showed that the intersection of human density and vectors per household produced the best estimates of epidemic potential (Ro per capita). Padmandabha et al noted that “when viral introduction is accounted for, human density amplified the effect of A. aegypti super-production on dengue risk”.  As they monitored the community over the summer with seven surveys they were able to see the decline in super-production decrease the epidemic potential in areas of highest human density.

These super-productive habitats (at the level of individual homes) are seen here to be critical in producing super spreading events of dengue fever. All of the parameters for what makes a super productive habitat including human behavior have not yet been fully explored. This study looked at residential areas with the same socio-economic status. This team is planning further studies that look at a range of socio-economic communities and incorporate community centers like schools and markets.  Studies like this one will be useful for designing strategies to target insecticide programs and other efforts to reduce mosquito abundance and dengue risk.

References
Padmanabha H, Durham D, Correa F, Diuk-Wasser M, & Galvani A (2012). The Interactive Roles of Aedes aegypti Super-Production and Human Density in Dengue Transmission. PLoS neglected tropical diseases, 6 (8) PMID: 22953017