Inbreeding and Disease in Tropical Shrimp Aquaculture: A Reappraisal and Caution27 May 2014
The disease crisis facing shrimp aquaculture may be propelled, in part, by an interaction between management practices that cause inbreeding, and the amplification by inbreeding of susceptibility to disease and environmental stresses. This study by Roger W Doyle, describes and numerically simulates gene flow from Penaeus (Litopenaeus) vannamei hatcheries that employ a ‘Breeder Lock’ to discourage use of their PL as breeders, through ‘copy hatcheries’ that breed the locked PL, to inbred shrimp in farm ponds.
Discussion of the current shrimp disease crisis focuses on the microbiology of pathogens and the regulations needed to limit their spread locally and across national boundaries (FAO 2008; Jones 2012; Lightner 2012; Reantaso 2012). In this study I take the perspective of a geneticist and evolutionary biologist, and suggest that a disease crisis that has microbiological roots may be amplified and accelerated by an agro-economic system in which shrimp breeders, hatcheries and farmers induce critical levels of inbreeding at farm level. The inbreeding is now manifesting itself as increasing susceptibility to disease and frequency of epidemics over vast areas of Asia, Central and South America, Africa and the Middle East.
The purpose of this study is neither to review comprehensively existing research on inbreeding in aquaculture nor disease problems in aquaculture, both of which are large and important topics. The purpose is more focused: to propose and, as far as possible with current information, to document a three-way connection between the effect of inbreeding in stressful conditions (especially disease stress), aquacultural practices that lead to inbreeding, and technical issues with widely used procedures for estimating inbreeding that have caused inbreeding to be underestimated. Many important papers that deal with aquacultural inbreeding or disease but do not bear on this three-way connection are not cited, for reasons of space and also to maintain focus. Because of these necessary omissions, the title of the study is entitled ‘reappraisal and caution’, not ‘review’.
The intended audience of the study includes technical specialists in any of the fields involved in this three-way connection: disease, genetics and aquacultural practice. Few will have expertise in all three, so the presentation attempts to be acceptable at a certain technical level and to remain comprehensible to non-experts.
The Results and Discussion section of the study is divided into five sub-sections: (1) description of a Breeder Lock aquaculture system that induces inbreeding in farmed shrimp populations, (2) numerical simulation of gene flow through this system and demonstration that batches of locked and copied postlarvae (PL) can be distinguished with appropriate use of genetic markers, (3) review of the effect of inbreeding on susceptibility to stress in general and shrimp disease in particular, (4) explanation of why we currently have so little information about inbreeding in shrimp hatcheries and farms and (mistakenly) believe inbreeding to be minimal, (5) possible policy initiatives that could reduce inbreeding and/or its effects.
Results and discussion
The aquaculture system that leads to inbreeding in farm shrimp populations
Broodstocks accumulate inbreeding and lose genetic diversity (correlated processes collectively called ‘genetic erosion’) when they experience bottlenecks or are chronically too small. A study of genetic erosion in wild and cultivated populations of Penaeus monodon in the Philippines (Xu, Primavera, Pena, Pettit, Belak & Alcivar-Warren, 2001) provides an exceptionally clear demonstration of this process, which is usually thought to be the main cause of genetic erosion in shrimp, i.e. (Perez-Enriquez, Hernández-Martínez & Cruza 2009). Broodstock managers are regularly warned about the problem and guidelines are repeatedly and widely disseminated (Tave 1993, 1995; FAO 2008, 2011). The mechanism underlying both components of genetic erosion is simple – random mating in a small population results in matings between biological relatives and random fixation of deleterious alleles. The main focus of this study, however, is a different cause of genetic erosion which is neither random nor directly related to population size: the ‘Breeder Lock.’
Gene flow from broodstocks, through hatcheries, to farms
Tropical shrimp farming comprises a transfer of genetic material, in the form of adult spawners, juvenile shrimp and postlarvae (PL), through a network of interconnected transactions between breeders, hatcheries and farmers. A major feature of these relationships – the ‘Breeder Lock’ and subsequent copying of locked animals – is described in detail in the following paragraphs because it is central to the proposed link between genetics and epidemics of shrimp disease. Fig. 1 shows the relationships in a diagram that describes the essence of the process in many parts of the world.
Spawners sent by breeders to hatcheries generally represent only a fraction of the total allelic diversity in the breeder's own broodstock (Gjedrem, Robinson & Rye 2012; Rye 2012), especially in well-managed programs where pedigrees are recorded; these will be referred to as ‘family breeding program’ (Gjedrem et al. 2012). Usually, the subset supplied to a hatchery comprises only two full-sib families of spawners, each containing thousands of brothers and sisters. The intention of the breeder is that the hatchery will generate production postlarvae (PPL) by mating these animals according to instructions that specify which spawners to mate together to produce high-quality offspring for distribution.
Production PL (PPL) flow onwards from hatcheries to farmers along two channels, only one of which is optimized by mating instructions from the breeders. Called here ‘authorized’ and shown as solid arrows in Fig. 1, PPL in this distribution channel are intended to have maximal uniformity and minimal inbreeding. The flow of genetic material along the solid arrows in Fig. 1 is similar to that recommended as good practice by Ponzoni, Nguyen, Khaw & Rodriguez 2012).
The ‘copy’ distribution channel shown as dashed arrows in Fig. 1 carries PPL that are diverted from the authorized channel – either by the hatchery or by farmers – and grown to maturity as broodstock in ‘copy hatcheries’. The offspring of these copy spawners will be inbred to varying degrees depending on the genetic composition of the authorized channel at the point where diversion takes place. Two successive bottlenecks occur in the gene flow from broodstock to farm shown in Fig. 1 (restricted subset of the broodstock sent to authorized hatchery, restricted subset of the PPL genome matured as copy breeders). Additionally, hatcheries tend to spawn as few shrimp as possible due to the fecundity of shrimp, and often use a highly skewed sex ratio because this enables them to maintain fewer brood shrimp (FAO 2008). Inbreeding in the copy channel is expected to range from 0.125 to 0.25 among PL in the first copy generation and as high as 0.375 in the next generation of random mating. Inbreeding resulting from the ‘Breeder Lock’ described below will be additive to prior inbreeding accumulated from previous locks or other processes of genetic erosion including small effective population size. Cumulative loss of genetic diversity over time and during transfers is well documented in shrimp (Benzie 2009; Perez-Enriquez et al. 2009).
The Breeder Lock that generates inbreeding
Although the primary objective of the authorized channel (solid arrows) in Fig. 1 is to provide highly uniform, non-inbred seed, an important secondary objective is to protect the breeder's intellectual property. Breeding programs are expensive, and breeders protect their investment in various ways, both contractual (Ogden & Weigel 2007) and biological. The most widely used biological defence against copiers is the one termed here the ‘Breeder Lock’, a mating scheme that produces highly inbred offspring in the copy channel. There are many possible ways to protect the intellectual property of broodstocks through genetic manipulation (Sellars & Preston 2008; Janhunen, Kause, Mäntysaari, Vehviläinen, Præbel, Järvisalo, Paananen & Koskinen 2013), but the simplest is probably the one described by (Doyle, Moss & Moss 2006) in which batches of many millions of seed animals are descended from just four grandparents, or four full-sib families of grandparents. A schematic diagram of the mating system that underlies this version of the Breeder Lock can be found in Doyle et al. (2006). This system is the basis of the simulation described in the 'Methods' section. PL from hatcheries that propagate seed according to instructions provided by the breeder normally give good results. However, seed produced by copiers are, as the breeders intend them to be, inbred and give poor growth and survival (Doyle et al. 2006; Sellars & Preston 2008; Gjedrem et al. 2012; Janhunen et al. 2013; Ponzoni et al. 2012; Argue, Tolentino & Brock 2014). The Breeder Lock in one form or another is widely used, defended and recommended for protecting the intellectual property of breeders, e.g. (Doyle et al. 2006; Ponzoni et al. 2012; Argue et al. 2014).
Copy hatcheries disseminate inbred shrimp
In the world of tropical aquaculture, an improved strain is generally copied shortly after it appears. Due to the high reproductive capacity of fish and crustaceans unauthorized reproduction and use of improved stocks tends to be widespread for many species (Rye 2012). Other hatcheries propagate the strain and sell later generations to farmers. They also mix inbred, copied animals with animals in the authorized distribution channel and sell the mix to unsuspecting farmers (dashed lines in Fig. 1). These activities have been grouped with poor broodstock management as ‘malpractice’ (Ponzoni et al. 2012).
Hatcheries in the copy channel may try to circumvent the locks by mating males and females from different hatcheries in the authorized channel but this tactic is probably futile. There are generally very few breeders – usually only one – supplying the authorized channel with non-inbred stock in any one aquacultural region (Ponzoni, Khaw, Nguyen & Hamzah 2010; Ponzoni et al. 2012). The breeder usually tries to supply all its client hatcheries with spawners from the same limited group of broodstock families.
Estimates of the global extent of copying
Most production of penaeids now depends on domesticated stocks (Stentiford, Neil, Peeler, Shields, Small, Flegel, Vlak, Jones, Morado, Moss, Lotz, Bartholomay, Behringer, Hauton & Lightner 2012). As Litopenaeus (Penaeus) vannamei is an exotic species in Asia, Africa and the middle East, it is entirely dependent on domesticated broodstocks in these regions.
Rye (Gjedrem et al. 2012; Rye 2012) estimates that production from uncontrolled breeding programs constitute more than 90% of worldwide hatchery production from all species. There is as yet no individual estimate for any shrimp species, but people directly involved in the tropical shrimp industry believe copying to be substantial. The following estimates have been offered as personal communications with permission to cite the source by name: Thailand, conservatively, 50% copied (Mr. Robins McIntosh); Ecuador >90% copied, Honduras ≈ 50% copied, Mexico >90% copied, Nicaragua ≈ 50% copied, Panama <10% copied, Venezuela >90% copied (Mr. José B. Martinez, Panama.). These estimates are in general agreement with consensus estimates developed during a recent international workshop on the possible connection between inbreeding and shrimp disease (NACA 2014).
The designation ‘copied’ in the preceding paragraph by no means implies that all copying involves a Breeder Lock. It also includes any broodstock that was initiated with restricted genetic diversity and propagated thereafter without a pedigreed family structure to limit inbreeding. Preliminary though it is, the information given above is all we currently have on the extent of copying in shrimp broodstocks. It may be taken as informative to an order of magnitude – that is to say, when properly estimated, production from copied broodstock is likely to be closer to 70% than 7% of the total.
Detection of the Breeder Lock-copy system: a simulation study
The question addressed in this section is not the global extent of copying and inbreeding, but whether PPL generated in a Breeder Lock can be recognized as such in the absence of pedigree information. To see whether this is technically possible requires comparison of the true inbreeding values (Fped) of broodstocks, production PL and copy PL with estimates based on genetic marker data. This is a technical question to which simulation offers a provisional answer. If the answer is ‘yes’, then practical tests can be devised for use at farm level.
Simulation has become a widely used procedure in population and quantitative genetics even in papers that are primarily analyses of observational data. Standard errors of the simulations rarely appear in purely theoretical studies because they are trivial functions of the parameters of the model and the number of simulation runs. In empirical studies where some aspect of a real-world set of data is simulated, standard errors, standard deviations or other measures of confidence are often provided to help assess the usefulness of the simulation. Two studies cited elsewhere in this study, (Whitlock 2002; Mattila, Duplouy, Kirjokangas, Lehtonen, Rastas & Hanski 2012), represent the first and second type of study respectively).
Results of simulation
Pedigree inbreeding Fped at all stages in the simulation is shown as solid bars in Fig. 2 and in tabular form in Table 1. Pedigree inbreeding in the hybrid, locked PPL (where Fped = 0) is statistically indistinguishable from that in the final generation of the broodstock (Fped = 0.002). Inbreeding in the copy PL (Fped = 0.168) is significantly higher than in broodstock and PPL (P < 0.0001). Breeder Locked production PL for distribution to farmers are recognizable by their large, negative and statistically significant Fis values, as shown in Table 1 and the cross-hatched and diagonally slashed bars in Fig. 2. The negative Fis statistic signals a deficit of homozygotes in the sample of 50 production PL relative to a hypothetical, random-mating population with the allele frequencies observed in the sample. This is a potentially easy and straight-forward way to detect the presence of a Breeder Lock. Neither of the TrioML estimates is significant in broodstock or production PL.
Both trioML and trioML(B) recognize the increa-sed inbreeding in the copy populations (0.05 > P > 0.01, P < 0.000, respectively; shown as diagonal-slashed and open bars in Fig. 2). The Fis estimator does not (P > 0.05), nor would it be expected to do so, as will be explained in section 'Current data and opinion concerning inbreeding in aquaculture'. TrioML does better than Fis in detecting inbreeding in copy PL when inbreeding and allele frequencies have to be estimated from the same samples, which is usually the case. However, if data from a meaningful reference sample are available – such as the founder population – trioML(B) should usually be preferred (third set of bars, Fig. 2). As Fis is the way to recognize the presence of Breeder Locked PPL, all three calculations are useful, in different contexts.
The simulation does indicate that a batch of genetically locked production PL can be recognized as such with co-dominant genetic markers (e.g. microsatellites) and an analytical technique (Fis estimation) that are widely used. Fig. 2 and Table 1 also indicate that samples from a broodstock – which is the source of most data in aquaculture genetic surveys – will not show the existence of a Breeder Lock. To do so requires that samples be taken from production PL. The sample size needed to detect the operation of a Breeder Lock with Fis at a specified level of statistical confidence will depend on the number of full-sib families in the reciprocal cross, allele diversity in the markers and other factors that have not been explored in the present simulation.
The trioML estimator uses more information about the population than the Fis estimator (Wang 2007) and Fig. 2 shows that trioML is able to recognize that inbreeding is higher in the first-generation copy PL than in the broodstock. Fis does not do so. Increased inbreeding is significant at the 0.05 level when the sample of copy PL is used as its own allele frequency reference (trioML), and at the 0.001 level when broodstock frequencies are used (trioML(B)). This is to be expected, as the calculated value of Fped should increase as the reference population is moved farther back in time and additional shared ancestors are included in the genealogy (Lacy 1995) while gene diversity also increases.
Inbreeding increases susceptibility to disease and other stresses
Interaction between inbreeding depression and environmental stress
Inbreeding depression is especially severe in environments where survival is low, even in outbred populations, owing to poor nutrition, extreme temperatures, the presence of pathogens or myriad other possible stressors alone or in combination (Frankham et al. 2002; Liao & Reed 2009; Cheptou & Donohue 2011, 2013; Bijlsma & Loeschcke 2012; Enders & Nunney 2012; Reed, Fox, Enders & Kristensen 2012).
Remarkably, data from many plant and animal taxa and many different kinds of natural and artificial stressors can be fitted to a common regression line of inbreeding depression against stress (Fox & Reed 2010). Their regression is shown as a dotted line in Fig. 3. There are actually three variables underlying the relationship in Fig. 3 although only two appear as axes in the graph. The Y coordinate of each point in Fig. 3 represents a coefficient, the effect of a causal variable (inbreeding) on a response variable (survival relative to a standard). Both variables are standardized to the same scale in all studies. The Y coordinates for the re-calculated OI data and the Fox and Reed data are therefore commensurate. The X-coordinates in Fig. 3, representing environmental stress, are also expressed relative to a standard common to all studies, i.e. non-inbred survival in an environment not exposed to the indicated stressor. Therefore, both axes are commensurate for the meta-analysis data and the re-calculated OI data.
Inbred and outbred experimental populations of the only shrimp in the Fox and Reed (2010) meta-analysis, the mysid shrimp Americamysis bahia, differed greatly in their survival under low salinity stress, and genetic load was much higher in stressful environments for several fitness indices (Markert, Champlin, Gutjahr-Gobell, Grear, Kuhn, McGreevy, Roth, Bagley & Nacci 2010). The authors note that this is actually an underestimate of the amplification of genetic load by stress because many of their inbred lines did not survive long enough to be tested. The fit of L.(P.) vannamei and other shrimp species to the meta-analysis regression of Fox and Reed (2010) cannot simply be assumed. This is true in general for new species and environment combinations, but especially for aquacultural species which, like L.(P.) vannamei and oysters, have exceptionally high fecundities and ‘huge’ inbreeding loads (Bierne, Beuzart, Vonau, Bonhomme & Bédier 2000). High fecundity may signal high sensitivity to inbreeding and it is also a reason why the Breeder Lock is easy to implement in shrimp. The two phenomena thus work together to increase the impact of inbreeding depression at farm level. Inbreeding depression in oysters, which have fecundities in the order of 105 offspring per spawn, has been studied in considerable detail (Bierne, Launey, Naciri-Graven & Bonhomme 1998; Launey & Hedgecock 2001; Plough 2012) and found to be high. The shrimp L.(P.) vannamei has a fecundity in the order of 105 offspring per spawn. Unfortunately, the amount of data is limited. While a number of excellent papers have been written on inbreeding in shrimp (e.g. Keys, Crocos, Burridge, Coman, Davis & Preston 2004). I cite here only those that include some type of experimental stress testing or comparison of inbreeding depression in different environments.
It should be noted that the molecular genetic cause of inbreeding depression is a long-standing, complicated and controversial issue in biology. The consensus, after several decades of research and theorizing, is that inbreeding depression is primarily caused by increased expression of recessive, deleterious alleles as the proportion of homozygous loci increases in a population and, to a lesser extent, by the parallel loss of expression of overdominant alleles (Whitlock 2002; Charlesworth & Willis 2009). This consensus now must be re-evaluated in the light of the recent discovery that a third fundamental process may be involved, epigenetic methylation (Vergeer, Wagemaker & Ouborg 2012; Cheptou & Donohue 2013). The effect of inbreeding on disease will probably turn out to involve some genes and epigenetic processes that are common to many pathogens as well as some that are pathogen-specific. A promising start on this type of genetic analysis has been made on oysters (Plough & Hedgecock 2011; Plough 2012) but not, as yet, on shrimp.
Inbreeding increases mortality from viral disease in L.(P.) vannamei
Two viruses, White Spot Syndrome Virus (WSSV) and Taura Syndrome Virus (TSV), bore most of the responsibility for the global economic loss from disease in shrimp as of 2011 (Stentiford et al. 2012). A new disease, Early Mortality Syndrome disease, EMS/AHPND, has recently become the most serious disease problem facing tropical shrimp aquaculture. EMS appears to be a septicaemic vibriosis involving at least two Vibrio species infected by a bacteriophage (FAO 2013a).
Litopenaeus (Penaeus) vannamei is by far the dominant shrimp species in aquaculture (Anderson & Valderrama 2013; FAO 2013b). Mortality induced by exposure to both of these viruses has been examined at various levels of inbreeding in a population of L.(P.) vannamei at the Oceanic Institute (OI), in Hawaii (Moss et al. 2007; Moss, Arce, Otoshi & Moss 2008). The linear relationship between inbreeding depression and environmental quality noted by (Moss et al. 2007) in disease challenge tests has been re-calculated as Ldiff to enable the OI data to be compared with those in the meta-analysis of Fox and Reed (2010). Fig. 3 shows that when a disease challenge test is treated as a stress like any other, the depressive effect of inbreeding on survival in L.(P.) vannamei is exceptionally strong relative to other environment-inbreeding interactions in the meta-analysis. This observation is cause for concern and may influence estimates of the threat posed by inbreeding when data on inbreeding at farm level become available.
The bootstrapped regression coefficient, estimated without a constant since Ldiff necessarily equals zero when stress equals zero, is statistically significant (P = 0.004). The regression of inbreeding load in L.(P.) vannamei with disease stress in Fig. 3 is significantly steeper than the meta-analysis of Fox and Reed (2010), which lies outside the 95% confidence limit of the L.(P.) vannamei regression.
High as it is in Fig. 3, the L.(P.) vannamei inbreeding loads under disease stress are comparable to loads in oysters (Bierne et al. 1998; Launey & Hedgecock 2001; Plough 2012).
An important and possibly unique field study has shown the effect of inbreeding mortality from disease in the farmed shrimp Litopenaeus stylirostris, in New Caledonia (Bierne et al. 2000; Goyard, Goarant, Ansquer, Brun, de Decker, Dufour, Galinié, Peignon, Pham, Vourey, Harache & Patrois 2008). Both components of yield, mortality and growth, were depressed by inbreeding that accumulated slowly over many generations (not rapidly, as in the Breeder Lock system described here). Inbreeding depression was especially evident in years when the environment was poor and overall yields low. This work is particularly relevant at this time because the disease affecting L. stylirostris was a septicaemia caused by a species of Vibrio, the bacterium which has recently been implicated (FAO 2013a) in the current EMS (or AHPND) disease crisis.
There appear to be only four studies that directly relate inbreeding to disease or any other stress in shrimp. Three are treated in some detail here: the data of Moss et al. (Moss et al. 2007, 2008) which are re-worked above, the L. stylirostris study in New Caledonia (Bierne et al. 2000; Goyard et al. 2008), and an inbreeding experiment on Penaeus monodon outlined by (Argue et al. 2014). All of these studies demonstrate a strong depressive effect of inbreeding on resistance to disease. There has also been one experimental inbreeding study on a domesticated stock of the shrimp Fenneropenaeus chinensis (Luo, Kong, Luan, Meng, Zhang & Wang 2014). Although the data on survival are reported in a non-standard way in the paper by Luo et al., it appears that when post-challenge survival in the non-inbred control group had dropped to 50%, survival in the most inbred group (F = 0.5) was 37%.
Other than shrimp, there are many aquatic organisms in the meta-analysis (Fox & Reed 2010). Shrimp are routinely challenged for a variety of diseases and other stresses in breeding programs. Usable data must therefore exist unexamined, or at any rate unpublished, in the files of many family breeding program that keep pedigree and mortality records.
Why non-inbred shrimp get sick, too
A common objection to the hypothesized importance of inbreeding as an amplifier of disease threats is that non-inbred animals from authorized hatcheries and wild populations get sick, too, and have always done so.
This observation, while true as narrative fact, is not relevant to the hypothesis that genetic erosion does not cause the disease crisis but makes it worse, possibly much worse. Animals in the Fox & Reed meta-analysis (Fox & Reed 2010) and papers cited therein did not die from inbreeding; they died from disease, cold, osmotic shock, poisoning, starvation, cannibalism and, perhaps, chagrin. But whatever the nature of the bad situation, inbreeding made it worse. An analogy may be helpful: inbreeding affect the severity and incidence of diseases analogously to the way drunk driving affects the severity and incidence of highway accidents. The observation that sober drivers have accidents, too, does not discount drunkenness as a threat to public safety.
The likelihood that an epidemic will break out is, even in highly structured populations, a positive function of susceptibility and contact rate between infected and non-infected individuals (Cross, Johnson, Lloyd-Smith & Getz 2007; Smieszek, Fiebig & Scholz 2009; Volz, Miller, Galvani & Meyers 2011). Inbreeding increases both factors. Thus, inbred (and, ex hypothesis, susceptible farms) can make a whole region more dangerous for all farmers, just as drunk drivers make highways more dangerous for everyone (Quinlan, Brewer, Siegel, Sleet, Mokdad, Shults & Flowers 2005).
Current data and opinion concerning inbreeding in aquaculture
The twofold objective of this study is to show that, (1) inbreeding at farm level likely to be widespread, and (2) inbreeding should not be ignored because it increases susceptibility to disease and other stressors. To make the second point, it seems necessary to show why inbreeding is being ignored now. Most recent reviews of diseases in aquaculture either conclude that inbreeding is of minimal importance (FAO 2008) or fail to mention inbreeding at all (Flegel 2012; Jones 2012; Lightner 2012; Moss, Moss, Arce, Lightner & Lotz 2012; Reantaso 2012; Stentiford et al. 2012; Chamberlain 2013; FAO 2013b; Murray 2013).
The consensus of most aquaculture scientists is that inbreeding will be a minor problem so long as established rules of broodstock management are followed. While correct, this consensus does not relate to the inbreeding on farms stocked from poorly managed broodstocks that may be the source of 90% of total production (Gjedrem et al. 2012; Rye 2012).
Another reason that non-specialists believe inbreeding to be minimally important is that field estimates from microsatellite markers are usually close to zero, and often negative. Moreover, most authors are careful to explain that, low as they are, such estimates are biased upwards by null alleles and unrecognized population substructure (Hedrick 2012) in the samples. The technical caveat concerning null alleles is often invoked to explain away an occasional statistically significant excess of homozygotes in samples, providing another reason for non-specialists to conclude that the impact of inbreeding is likely to be small.
Geneticists are rather less careful to point out that most microsatellite estimates actually provide no direct information whatsoever about inbreeding in aquacultural broodstocks. There are two, related difficulties with these estimates. Firstly, the indicators of inbreeding most often reported are deviation from Hardy–Weinberg equilibrium and/or the fixation index, Fis. And secondly, amova and simpler procedures for estimating H-W and Fis are usually based on allele frequencies in the same set of samples for which the estimate is made. These difficulties are sufficient to invalidate most, but by no means all published marker-based estimates of inbreeding in aquaculture. (The L. stylirostris study in New Caledonia (Bierne et al. 2000) is a notable exception.)
Fis is a poor indicator of inbreeding
The fixation index, Fis, is an indicator of non-randomness in the mating system and thus indicates a potential cause of inbreeding but does not directly measure inbreeding (Templeton & Read 1994). Deliberate non-random mating is rare in aquaculture except in the Breeder Lock situation shown in Fig. 1. Instead, inbreeding in aquaculture broodstocks nearly always accumulates owing to bottlenecks, small population sizes, unequal fecundity and other random processes, rather than deliberate consanguineous mating. ‘In finite populations some individuals mate with biological relatives just by chance and inbreeding in the pedigree sense is the result’ (Templeton & Read 1994). The resulting inbreeding will not produce a significantly positive Fis so long as mating is random. Even the offspring of a population of full-sib brothers and sisters, where Fped = 0.25, shows neither H-W deviation nor positive Fis if the siblings mate at random.
Furthermore, the fixation index Fis (and deviation from H-W equilibrium) are signals of non-random mating in the parental generation immediately preceding the generation providing the genotype and allele frequencies. As it is based on allele frequencies, not genotype frequencies, Fis loses all information about accumulated inbreeding, if any exists, after a single round of random mating. This is seen in Fig. 2, where the only Fis differing from zero is the PPL stage, which comprises offspring of forced hybridization (reciprocal cross). The non-random mating of broodstock in a Breeder Lock thereby produces an Fis value that becomes a signature of the lock. Inbreeding in the copy stage of Fig. 1 is caused by random mating of PPL in copy hatcheries, and the Fis estimate in their offspring drops to zero even though Fped has risen to 0.136 (Table 1).
The second difficulty with Fis and other indicators of inbreeding/non-random mating arises when allele frequencies are calculated from the samples for which estimates are to be made. ‘Fis is defined with respect to the populations that are included in the sample, either through population-specific estimates or through the average of those estimates.’ (Holsinger & Weir 2009). The practical and conceptual difficulty arising from this is beautifully explained by Wang (2014), whose paper should be studied by anyone seeking further information. ‘Frequently genotype or allele frequency data are unavailable from an ancestral population, and allele frequencies used in calculating relatedness have to be estimated from the current sample in which relatedness between individuals is being calculated. This practice effectively uses the current population (sample) as reference, and an estimator conforming to the correlation concept of relatedness should give an average estimate of zero. This is true regardless of the actual relatedness among individuals in the sample, as shown by simulation and analytical results in (Wang's 2014) study.'
The following general summary applies with special force to aquaculture: ‘Fixation indices (Fis, FST…) allow to quantify inbreeding at the population level with respect to non-random mating…. no such measure exists for inbreeding due to small population size or bottlenecks.’(Szulkin, Bierne & David 2010) [emphasis added]. This fact may surprise non-specialists who believe that inbreeding is unimportant when in fact we have hardly any direct evidence concerning inbreeding in hatcheries that lack pedigree data.
The obvious solution to both difficulties is to use maximum likelihood estimators such as Wang's trioML rather than Fis, and, if possible, base inbreeding estimates on reference allele frequencies from an earlier generation, as in the trioML(B) estimator of Fig. 2. An appropriate reference is the generation that would be called the founder of the pedigree had pedigree records been kept. A survey of L.(P.) vannamei broodstocks in Mexico (Perez-Enriquez et al. 2009) is a good example of this approach. For their study of Litopenaeus stylirostris in New Caledonia (Bierne et al. 2000) used published data from wild populations of other penaeid species – a bold move.
Possible ways to recognize, monitor and reduce inbreeding
Farmers should be able to obtain verifiable information about the inbreeding status of PL offered to them for sale or already stocked in their ponds. Pedigree information is rarely available from PL suppliers in the real world of shrimp farming, which involves transfers of animals among groups of people with very different commercial objectives and record-keeping procedures, over a period of years.
Verifiable certificates of authenticity might persuade farmers to avoid the copy channel
The consensus of breeders is that farmers ought to learn the consequences of buying copied PL the hard way, which they do. Farmers are aware of inbreeding depression, however, and often ascribe their disease problems to it (FAO 2008). They also have a good notion of how broodstocks are managed by copy hatcheries in their local areas. However, farmers usually cannot be sure the seed animals they purchase are not inbred even when they buy from supposedly authorized hatcheries. Their puzzling reluctance to pay more for genetically superior aquaculture stock (Ponzoni, Nguyen & Khaw 2009; Gjedrem et al. 2012) may be due in part to lack of credible information. This possibility has already been noted (Ponzoni et al. 2012).
If they were offered verifiable information by authorized breeders and hatcheries, farmers could, if they chose to do so, avoid the copy distribution channel in Fig. 1. ‘Certificates of authenticity’ have been offered by some breeders, but this strategy fails when the certificates too are copied. Certificates offered to date have been missing the essential element of verifiability. Verification of the location of a population in Fig. 1 is technically easy in principle. Certificate from breeders attesting that a particular batch is 100% heterozygous for two particular alleles at a particular locus (both specified in the certificate) would be sufficient to verify that the batch is a first-generation PPL hybrid and minimally inbred. If certificates of authenticity are unavailable, the simulation of the lock & copy mechanism in Fig. 1 shows that locked (i.e. hybrid) production PL can be recognized and distinguished from copy PL by their strongly and significantly negative Fis.
Inbreeding at farm level can be estimated
Estimating the amount of inbreeding which has already accumulated in tropical shrimp hatcheries and farms will require modification of the statistical procedures that are currently most often used. The modification suggested here is to use Wang's trioML estimator and refer the analysis of samples from farm ponds back to an appropriate reference population and generation, e.g. as in trioML(B). Using samples from ponds or broodstocks as a self-reference for frequencies simply does not estimate accumulated inbreeding.
Transboundary restrictions, disease resistant and SPF stocks, and the lock-copy system
International organizations concerned with aquaculture, including the Food and Agriculture Organization of the United Nations (FAO), are responding vigorously to the disease crisis by developing regulations and codes of practice for transferring aquacultural stocks between and within regions, and also by promoting standard and guidelines for disease detection, identification and surveillance (FAO 2008; Jones 2012; Ponzoni et al. 2012; Reantaso 2012). The culture of stocks certified to be free of specified pathogens (SPF stocks) is strongly recommended and only SPF stocks can now be legally imported into most jurisdictions. These recommendations are appropriate, beneficial and necessary from a microbiological perspective. But insofar as they increase the value of proprietary, high-quality SPF strains such regulations may also increase the use of the Breeder Lock and the likelihood of copying, and thus inbreeding at farm level.Intellectual property rights are fundamental to science-based economic innovation. Breeders will, and should, continue to protect their genetic improvement programs with the Breeder Locks that generate inbreeding when copied, especially in regions where judicial sanctions are ineffective. The intellectual property value of genetically disease resistant strains will be extremely high. The regulatory objective should be to encourage biosecurity and genetic progress while at the same time discouraging copying and consequent inbreeding. The Network of Aquaculture Centres in Asia Pacific has made the first move in this direction (NACA 2014).
In this study, I suggest that disease crises in tropical shrimp aquaculture may be growing more severe and more frequent owing to an agro-economic system that generates genetic erosion at farm level. Genetic erosion results from a pattern of human behaviour in which breeders protect intellectual property through the Breeder Lock (expressed only when broodstock is ‘copied’), copying hatcheries sell inbred offspring, and farmers stock their ponds with seed animals they are unable to evaluate. The resulting inbreeding and low genotypic diversity increase susceptibility to disease, which leads to more infected individuals and farms and thus, by standard epidemiological reasoning, increases the frequency and severity of epidemics.
The hypothesis is not that inbreeding triggers shrimp diseases – which have myriad environmental and other immediate causes – but that inbreeding increases the prevalence and severity of disease, and that inbreeding is accumulating on regional scales. We may be making a mistake in treating the torrent of shrimp diseases of the past decade as isolated events with independent, microbiological causes, describable with some (unknown but invariant) distribution of risk. The distribution of risk may be moving towards higher values in a farming system experiencing genetic erosion.
The case made here for the role of genetic erosion in the disease crisis is both circumstantial and prima facie, meaning that while this study may succeed in showing that that there is a case to answer, the case is far from conclusive. Information that would settle the matter – the extent of broodstock copying, the true level of inbreeding in farms and the magnitude of inbreeding depression in shrimp may be obtainable by relatively simple adjustments to current procedures and re-interpretation of existing data.
You can view the full report by clicking here.