Haydon et al.’s (2002) “reservoirs framework” paper provides a structure for understanding reservoirs of infection by distinguishing between maintenance, source and target populations and clarifying the relationship between them (see Viana et al. 2014 Box 1, Figure I for an open access explanation). Reading it for the first time a few years ago, I found myself drawn into testing the structures and the relationships between them, generating new examples, verifying that the framework provided was sufficient to describe them, and checking whether they were topologically equivalent to those depicted in the figure. It felt like learning a new axiomatic system in mathematics. At the time, I paid very little attention to the terminology. The language of squares, circles and arrows seemed sufficient.
The Black Death, probably caused by Yersinia Pestis and depicted here on a panel of the Great Tapestry of Scotland, has long been one of the most iconic examples of a pandemic implicating a reservoir of infection and explanations of its mechanisms still attract considerable scientific interest. Photo: Alex Hewitt/Trustees of the Great Tapestry of Scotland (GTOS)
Although our recent update (Viana et al., 2014) focused primarily on reviewing threads of evidence and discussing ways in which they might be woven into a tapestry to allow us to better identify reservoir systems, its writing has also led to new discussions about the framework and the structural relationships it encompasses. But it has also led to discussions of terminology. Most of these have focused on our use of the term ‘reservoir’. I’ve spent the last couple of hours trawling online dictionaries and etymological sources in order to better understand our use of the word. A foray into another of my favourite worlds: language. Understanding the different meanings of the word might also help us to understand some of the uncertainties that have been raised about the framework.
Most free online English language resources, such as Merriam-Webster and Online Etymology Dictionary, provide fairly limited information on the etymology of the word reservoir beyond a reference to its French origins. The Centre National de Ressources Textuelles et Lexicales, a CNRS resource centre, helps us to trace the term a little further back. According to the entry in their etymological dictionary, the first recorded use of the term dates back to 1510 when it was used to refer to a receptacle for holding a liquid. By 1547, it was being used more generally as a space fitted out for the conservation and storage of provisions, and by 1601 had adopted a figurative meaning, being used to refer to anything capable of serving as a repository. Despite these subtle changes, these uses all relate to the notion of a container employed for purposeful storage. Perhaps surprisingly, it was only in 1742 (in French; and slightly earlier in English according to the OED) that it took on the meaning of a place serving as a natural reserve of something. Yet among modern definitions, even if we exclude epidemiological meanings, we find a third use of the term as a supply of something in which the reservoir no longer refers to the container but to a resource that is contained. I suspect this is the sense in which Acheson employed it in explaining that Winston Churchill
“… still had his glorious sense of words drawn from the special reservoir from which Lincoln also drew, fed by Shakespeare and those Tudor critics who wrote the first Prayer Book of Edward VI and their Jacobean successors who translated the Bible.”
Dean Gooderham Acheson (1961) Of Winston Churchill in Sketches from Life of Men I Have Known.
Actually, what alerted me to the different meanings was not the etymology at all, but prepositions, something else that Churchill is reputed to have been sensitive to. What none of the above sources note is that the distinction between the different meanings of the word reservoir can be detected in its association with particular prepositions. When using it in the sense of a purposeful receptacle, we use the preposition for, such as when we refer to a ‘reservoir for heating oil’; in the sense of a natural reservoir, we would generally employ the preposition of, as we might if talking about a ‘subterranean reservoir of natural gas’. In the case where we want to emphasise the idea that a natural reservoir serves as a supply, we use both prepositions, but the meaning of the word for now changes. In the phrase ‘a subterranean reservoir of natural gas for the population of Scotland’, the word for refers not to the natural gas (as it did in the heating oil example) but to the population due to receive the gas.
In the epidemiological context, the equivalent of a reservoir of natural gas for a population would look something like
A reservoir of [infectious agent] for [target population].
And yet, the epidemiological literature is replete with examples of the equivalent of ‘a reservoir for natural gas’ (i.e. a receptacle into which one puts natural gas). A search in my Mendeley library brings up a list of examples: ‘a potential reservoir for Leishmania’, ‘a reservoir for a coronavirus’, ‘the reservoir for the origin of the SARS epidemic’, ‘a reservoir for emerging infectious diseases’, ‘a reservoir for rabies’ and ‘a reservoir for bovine tuberculosis’. When we write in this way, I am sure that we are simply being imprecise rather than implying a sense of human purpose in the maintenance of these reservoirs. But we really should try to use language a bit better than that.
But how does this distinction relate to the question of how we describe structures using the reservoir framework? Firstly, it explains why we choose to refer to the target in the definition of a reservoir. Basing our definition on that of Haydon et al. (2002), we explain that “A ‘reservoir of infection’ is defined with respect to a target population as ‘one or more epidemiologically connected populations or environments in which a pathogen can be permanently maintained and from which infection is transmitted to the target population’”. Thus, according to the framework, referring to a reservoir without reference to a target constitutes under-specification. Obviously, without maintenance there would be no reservoir; but equally, if there were no target population into which disease spills over then the term maintenance population would fully characterise the system and there would be no need to refer to a reservoir. For example, for a multi-host pathogen such as the virus causing foot-and-mouth disease, referring to buffalo as ‘the reservoir’ makes little sense because the system is under-specified: if we complete the definition by specifying a target, the factual accuracy of the statement “buffalo are the reservoir of FMDV for <target>” depends on the particular target we choose.
More precisely, the framework in Haydon et al. (2002) should be thought of as serving to describe not just reservoirs, but target-reservoir systems. According to the framework, populations and communities are classified in two ways. Firstly, according to their maintenance status as either capable of maintaining the pathogen in the long term or incapable of doing so; and secondly, according to their role in transmission between populations within the target-reservoir system as target, source, or neither. The simplest way to characterise the full system is then to view these dimensions as orthogonal: every population has an attribute from each of the two dimensions.
This construction helps to answer a number of questions that have arisen in discussion with colleagues. For example, it means we may still wish to refer to a reservoir even if the target population is capable of maintenance (or R0 in the target is greater than one). For example, this would be the case if some infections in the target came from other maintenance populations in the system. Furthermore, a source population can be maintenance or non-maintenance. Source populations that are not capable of maintaining a pathogen alone can form an essential or inessential part of a maintenance community, or simply assist in the transfer from the maintenance population to the target. In fact, all three possibilities might be involved in the transmission and persistence of the plague bacterium, Yersinia Pestis, in relation to different flea species and mammalian host communities (Eisen & Gage, 2009; Webb, Brooks, Gage, & Antolin, 2006). One might ask, as Ashford (2003) has, whether or not vectors that do not contribute to pathogen maintenance should be included in the reservoir. As Ashford notes, this particular point could be argued either way; nonetheless, distinguishing between types of vectors is important when designing interventions.
Fundamentally, there are two ways to protect the target: either we prevent maintenance, or we prevent transmission from the maintenance community to the target. As we explain in Viana et al. (2014), there are various ways to achieve these aims, which we refer to as press, pulse and block. However, this categorisation focuses on the implementation rather than the aim. For example, a pulse intervention may consist of culling (to prevent maintenance) or vaccination (to prevent transmission to the target); a block action may employ fences erected between non-target, non-maintenance populations (to prevent community-level maintenance) or between a maintenance community and the target (to prevent transmission to the target).
Figure 1. Simple reservoir-target systems showing the three kinds of vectors. T denotes the target population, V the vector source and P an additional population involved in the target-reservoir system. Arrows indicate transmission between populations, circles represent non-maintenance populations while squares are maintenance populations; maintenance communities are shown with a dashed outline.
To come back to the importance of distinguishing between vectors that are involved in maintenance and those that are not, an interesting case arises. In Figure 1, although eliminating the vector is effective for different reasons in the three cases, the set of interventions is actually identical (eliminate population P, eliminate the vector population V, block transmission link a, block transmission link b). In the first and third case, eliminating V is effective because it breaks the transmission link to the target; in the second case, its elimination is also prevents maintenance in the community consisting of the vector V2 and population P2.
Ultimately, perhaps the most important questions about the definition and associated framework relate not to the word reservoir in the definition, but to how generally the framework applies. For example, should we use it for situations such as environmental persistence without pathogen reproduction? It seems fairly natural to apply it to when considering to parasites, but should it extend to organisms such as toxin-producing algae or fungi that do not require living matter in order to reproduce? These are all fascinating questions and it should be fun thinking about whether and how to best integrate them.
(With thanks to Daniel Haydon and Mafalda Viana for comments.)
[RRK – Comments can be made here, or addressed directly to rebecca.mancy A T glasgow.ac.uk]
Ashford, R. W. (2003). When Is a Reservoir Not a Reservoir? Emerging Infectious Diseases, 9(11), 1495–1496.
Eisen, R. J., & Gage, K. L. (2009). Review article Adaptive strategies of Yersinia pestis to persist during inter-epizootic and epizootic periods. Veterinary Research, 40(1), 1–14.
Haydon, D. T., Cleaveland, S., Taylor, L. H., & Laurenson, M. K. (2002). Identifying reservoirs of infection: a conceptual and practical challenge. Emerging Infectious Diseases, 8(12), 1468–73. Retrieved from http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2738515&tool=pmcentrez&rendertype=abstract
Viana, M., Mancy, R., Biek, R., Cleaveland, S., Cross, P. C., Lloyd-Smith, J. O., & Haydon, D. T. (2014). Assembling evidence for identifying reservoirs of infection. Trends in Ecology & Evolution, 29(5), 270–279. doi:10.1016/j.tree.2014.03.002
Webb, C. T., Brooks, C. P., Gage, K. L., & Antolin, M. F. (2006). Classic flea-borne transmission does not drive plague epizootics in prairie dogs. Proceedings of the National Academy of Sciences of the United States of America, 103(16), 6236–41. doi:10.1073/pnas.0510090103