The future of predictive microbiology: Strategic ...

5 downloads 12239 Views 251KB Size Report
is now also readily available for use on laptop computers to assist in the development of .... lems with data recovery, retrospective analysis of information and.
International Journal of Food Microbiology 128 (2008) 2–9

Contents lists available at ScienceDirect

International Journal of Food Microbiology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / i j f o o d m i c r o

The future of predictive microbiology: Strategic research, innovative applications and great expectations Tom McMeekin ⁎, John Bowman, Olivia McQuestin, Lyndal Mellefont, Tom Ross, Mark Tamplin Food Safety Centre, School of Agricultural Science and Tasmanian Institute of Agricultural Research, University of Tasmania, Private Bag 54, Hobart, Tasmania, Australia 7001

A R T I C L E

I N F O

Article history: Received 1 February 2008 Received in revised form 22 May 2008 Accepted 29 June 2008 Keywords: Predictive microbiology Model building Strategic research Enabling technology Value analysis Modelling food and other ecosystems Microbial persistence and recovery

A B S T R A C T This paper considers the future of predictive microbiology by exploring the balance that exists between science, applications and expectations. Attention is drawn to the development of predictive microbiology as a subdiscipline of food microbiology and of technologies that are required for its applications, including a recently developed biological indicator. As we move into the era of systems biology, in which physiological and molecular information will be increasingly available for incorporation into models, predictive microbiologists will be faced with new experimental and data handling challenges. Overcoming these hurdles may be assisted by interacting with microbiologists and mathematicians developing models to describe the microbial role in ecosystems other than food. Coupled with a commitment to maintain strategic research, as well as to develop innovative technologies, the future of predictive microbiology looks set to fulfil “great expectations”. © 2008 Elsevier B.V. All rights reserved.

1. Introduction 1.1. Strategic research, innovative applications and great expectations in the context of predictive microbiology The Australian Research Council (www.arc.gov.au) defines “oriented strategic basic research” (herein abbreviated to “strategic research”) as “research carried out with the expectation that it will produce a broad base of knowledge likely to form the background to the solution of recognised or expected current or future problems or possibilities”. The key term, “innovative applications”, is defined as “the use to which something, new or improved, may be put, for example, scientific knowledge, especially in industry, and equipment for the purpose”. These were chosen as elements of the title to indicate that a combination of science and technology is essential for effective application of predictive microbiology (McMeekin et al., 2005). The key term, “expectations”, is defined as “the act of looking forward, something hoped for” or (as in Charles Dickens' novel Great Expectations “prospects for inheritance”) and, as such, represents the practical outcomes of predictive microbiology research. This is very relevant to current consumer expectations of food, as clearly stated by Carol Brookins at the Global Food and Agriculture Summit in 1999:

⁎ Corresponding author. Tel.: +61 3 62266280; fax: +61 3 62267444. E-mail address: [email protected] (T. McMeekin). 0168-1605/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ijfoodmicro.2008.06.026

“Consumers are demanding miracle foods that are totally natural, have zero calories, zero fats and cholesterol, delicious taste, total nutrition, low price, environmentally friendly production, ‘green’ packaging …. and that guarantee perfect bodies, romance and immortality.” This is even more relevant in 2008. A related term, “expected value”, is “the predicted value of a variable calculated as all probable values multiplied by the probability of its occurrence”. Clearly, the concurrence of expected and measured values provides an estimate of the value of a predictive model or “a measure of its worth, desirability and utility”. Quantifying the value of predictive microbiology research has rarely, if ever, been attempted before, in other than a rudimentary manner, but herein we will describe the outcomes of a value analysis—“a systematic and critical analysis of a process or every feature of a product”—carried out by the Centre for International Economics, Canberra, Australia, for Meat and Livestock Australia (www.mla.com.au). 1.2. The mindset engendered by predictive microbiology In several earlier publications we have drawn attention to the writings of Scott (1937) as the first clear enunciation of the concept of predictive microbiology and, in the context of this paper, it seems reasonable to mark this as an example of creative and innovative thinking, raising expectations of changing agar-based approaches to microbial enumeration and, as a result, enabling proactive rather than retrospective estimation of the microbial safety and quality of foods.

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9

However, without appropriate technology such as data loggers, computers and the internet, Scott’s concept was destined to remain “virtual” until those innovations were developed and had become readily available. Thus, the era of “modern” predictive microbiology can be traced from the early 1960s when the science and the technology required for predictive microbiology came into balance. Despite the fact that Scott’s predictive modelling concept reported in 1937 was not a practical reality until 30–40 years later, the gestation period should have allowed early proponents to move towards an understanding of the potential impact of the new approach. Expectations of the transition from qualitative to quantitative microbial ecology of foods would have included reduced uncertainty, possibilities becoming probabilities and variability in response times characterised by appropriate distributions (McMeekin, 2007). These effects were already well understood by insurance companies, which used the Gompertz model (Gompertz, 1825) to predict life expectancy (Gross and Clark, 1975; Teriokhin et al., 2004; Bongaarts, 2005). Stock market analysts whose life (or work) expectancy is likely to be highly correlated with volatility in the market, will also be acutely aware of uncertainty. Variability is a biological reality described by a distribution which allows its magnitude to be estimated, but not altered, for any set of conditions. Uncertainty, on the other hand, can be reduced by collection and analysis of more data, but it always retains the potential to be activated, often as a result of human error. A more practical prospect from the introduction of enabling technologies, that would have grabbed the imagination of early predictive microbiologists, was the expectation that estimates of shelf life and safety would be available in a compressed time frame; although the prospect of real-time reporting of food safety and shelf life in the 1960s may have been too large a leap of faith to comprehend. 2. Why model, who models and how are models built? A model can be defined as “the description of a system, theory, or phenomenon that accounts for its known or inferred properties and may be used for further study of its characteristics”. While, in common usage, a model is often a smaller replica of a real object, in science, engineering, finance etc., the model is an often simplified description of relationships between observations of the system (responses) and the factors that are believed to cause the observed responses. That description can be expressed in words or expressed quantitatively in one or more mathematical relationships or equations. Thus, a mathematical model can simply describe a collection of data or may represent a hypothesis or series of hypotheses about underlying relationships among the independent variables that lead to the observations or data.1 The first approach is often termed an ‘empirical’ model, while the latter is described as ‘mechanistic’. Both approaches have utility: the first simply to summarise data and the latter to summarise “understanding” or knowledge. Either can be used to predict the response of the system to changes in the variables. Few models are truly mechanistic, but models can at least be formulated to reflect and embody our current knowledge and/or hypotheses concerning the system being studied. In this way, predictions from the model can have utility not only to predict the outcomes of a set of circumstances, but also to test the hypotheses embodied in the model and so to revise and improve the model and our hypotheses if the predictions do not match the observations. This is the “traditional” scientific method. Thus, models not only provide a framework to summarise our experience about what will happen under a given set of circumstances, but also a systematic way to improve our understanding of the underlying processes. With that understanding we are better able to predict, or control, or improve the performance of a system intelligently, whether it be the stock market, 1 A series of mathematical equations that are used to solve a problem (usually involving repetition of one or more operations) can also be called an algorithm.

3

traffic flow, agricultural soil fertility, global weather or the microbiological stability of foods. Specific approaches to model building will differ from discipline to discipline and even within disciplines, depending on the system being modelled. Nonetheless, all models seek to link observations to the variables believed to control them so that all models require quantitative data on the magnitude of responses and the variables believed to influence them. Mathematical techniques to summarise or identify the relationships between independent and dependant variables are well established and documented in many texts, as are principles of experimental design to generate the needed data and discern the relationships. Issues of relevance include the range of variables studied and the limits of applicability of model predictions, correct specification of error behaviour (the “stochastic assumption”), model parameterisation and parsimony (not having too many explanatory variables), etc. (see Ratkowsky, 1993). A wide range of powerful regression software tools is now also readily available for use on laptop computers to assist in the development of reliable and robust mathematical models. Access to such powerful software has increased the use of modelling approaches to a wide range of human endeavours. 3. Predictive models: enabling science Ample evidence is provided in the history of science to demonstrate that advances occur from making observations as a basis for hypotheses on which general rules describing natural phenomena are constructed. In food microbiology, qualitative approaches to describe microbial population responses predominated until the advent of “modern” predictive microbiology which allowed us to advance to the era of the quantitative microbial ecology of foods. In turn, the phenomenological descriptions may be strengthened by suppositions, based on initial facts, leading to better defined hypotheses and, perhaps, to mechanistic insights and underlying theories to characterise the phenomena involved. In predictive microbiology, most models used to date are empirical in nature ranging from “black box” approaches, such as artificial neural networks, to “grey box” models which include a priori knowledge to describe well characterised microbial responses to environmental factors (Geeraerd et al., 2004). Progress towards a thermodynamically based temperature dependence model was reported by Ratkowsky et al. (2005) and this has moved our state of knowledge closer to a mechanistic basis for temperature effects on microbial growth rates, based on reversible protein denaturation at both low and high temperatures. Similarly, models of the effect of other constraints, such as pH and water activity, on microbial growth rate, when combined with physiological studies, provide mechanistic clues. An example is the severe drain on energy reserves through induction of ATPase to expel protons from cells in a low pH environment, thus markedly reducing cell yield at decreasing pH levels (Krist et al.,1998). The model form may also provide evidence to describe the nature of combined effects on microbial growth, the core of the classical hurdle concept introduced by Leistner and colleagues as the basis of “mild” processes for food preservation [see Leistner and Gorris (1995) and Leistner (2000) for developments in application of the concept]. The prevailing consensus is that hurdles interact additively in reducing growth rate rather than in much sought after synergisms (Lambert and Bidlas, 2007). The preamble above on models is provided to stake a claim for the central role of strategic research in predictive microbiology. Before the advent of quantitative microbial ecology approaches, we relied heavily on qualitative or semi-quantitative approaches, usually applied in discrete experiments, to specific situations. For example, “The effect of environmental conditions A, B and C (or antimicrobials D, E and F) on the growth and inactivation of microorganisms P, Q and R in foods processed and stored under conditions X, Y and Z” could be a generic title applicable to many publications in the recent food microbiology history. Analysis of non-review paper titles published in the Journal of

4

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9

Food Protection and the International Journal of Food Microbiology in 2007 reveal that our generic title is applicable to ~14% and 5.5%, respectively. Effectively, this approach confirms that empiricism continues to be used in the search to uncover combinations of control strategies with benign effects on the sensory or nutritional properties of the food and the health and well being of consumers. Serendipity is well recognised as important in scientific advances, but the opportunity it offers needs to be recognised and, therefore, should be considered as an observational starting point from which mechanisms and applications are developed. Systematically “filling in the dots” between empirical and mechanistic descriptions is a longer, but more certain, route to secure scientific advances. The innovation arising from predictive models is that they allow the interpretation of the effect of processing, distribution and storage practices on the extent of growth or death of the organism for which the model was developed, i.e. microbial behaviour is predictable on the basis of measured environmental parameters. Additionally, one environmental history profile may be applied to several organisms, if appropriate models are available. With addition of the modelling innovation, the proposed generic title of our phantom paper would change to “Mathematical models describing the effect of environmental conditions A, B and C (or antimicrobials D, E and F) on the growth and inactivation of microorganisms P, Q and R in foods processed and stored under conditions X, Y and Z”. However, the quantitative approach would require much greater experimental effort and analysis with attention to the “rules” of modelling outlined in Section 2 and described in detail by Ratkowsky (1993), experimental design, protocols, etc. (McMeekin et al., 1993). The process of model building is not simply a matter of finding an equation to describe a single or a few data sets. In fact, the pattern of microbial responses observed in relation to environmental variables is likely to provide more useful information than a curve fitting exercise which will provide a unique solution with little, or no, utility beyond the experimental conditions under which the data were generated. Confidence in predictions is greatly increased when models from different sources provide comparable estimates of environmental influences on microbial population behaviour. However, as with Scott (1937), practical application of models in foods still requires technological “fixes” which we will consider in the following section. On the other side of strategic modelling science we encounter the interface with microbial physiology (“-omics” in current parlance), the priorities for the study of which are often set on the basis of modelling studies, e.g. the fascinating region close to the growth/no growth boundary (Ratkowsky and Ross, 1995; McMeekin et al., 2002). From physiological and molecular studies, one can predict both the identification of specific targets for control of microorganisms and predictive markers to identify the onset of significant biological events such as sporulation and germination (Oomes et al., 2007). As the emergence of modern predictive microbiology depended on the invention of computers, identifying the mechanisms of physiological responses to environmental stimuli (i.e. systems biology) requires sophisticated computational tools to sort through huge knowledge databases and “map” complex networks. Fortunately, such tools are rapidly emerging as a result of accelerated research in network sciences, a discipline that describes networks which are comprised of links (edges), nodes and “hubs” that influence responses in virtually all systems [see Barabasi (2002)]. 4. Technological innovations: enabling devices Before considering types of devices to monitor environmental conditions, it is important to understand the pivotal role of databases such as ComBase (www.combase.cc) which allow us to “join the dots” or “network”, thereby moving from discrete data sets useful in a standalone situation, possibly compared with similar data obtained by browsing the literature, to a much more knowledge powerful compen-

dium of stored information to which new data may be added going forward. Importantly, much information is made freely available to users through such databases which are a virtual science network, geared to the refinement and dissemination of knowledge and increasingly obviate the need for reinvention of expensive experimental wheels. 4.1. Traceability technologies Traceability technology, such as barcoding or radio frequency identification technology (RFID), allows a product to be followed step by step or to be recognised as an object to which an original and unique code was applied. This authenticates that it is genuine or has an undisputed origin, provided that illegal product substitution has not occurred. Traceability and authenticity technologies may be as simple as a “one up, one down” paper trail by which each business operator can identify their supplier and customer and, on demand, provide this information to the competent authorities (McMeekin et al., 2006a). An early example of authenticity being confirmed by a pen and paper system was used for N100 years by the American Waltham Watch Company to record the details of millions of timepieces against a unique serial number stamped in the mechanism. This would have been useful for specialist access by owners and collectors wishing to authenticate the potential value of their horological devices, but now that the information is digitised, general access has been greatly increased (see www.walthamwatchcompany.com). For most consumers, barcodes, machine readable codes in the form of a pattern of stripes printed on and identifying a commodity, especially for stock control, will be the most commonly recognised traceability technology. Variants on barcodes include information dense microbial dots (www.burntsidepartners.com) and the fortuitous dots code described by Neimeyer-Stein (2006). Electronic chain distribution monitoring systems, many based on RFID technology, have attracted significant attention in the quest for advances in product monitoring in the logistics industry. In addition to product identification via barcodes, many systems also allow temperature measurements to be transmitted in real-time at any point in the supply chain (Frederiksen et al., 2002). Thus, the addition of another function has increased the utility of the system by allowing identification of weak points in the cold chain. An important consideration is the need for complementarity of RFID technologies through common standards and a common identifier, the Global Trade Item Number (see www.epcglobalinc.org) (McMeekin et al., 2006a). 4.2. Temperature function integration (TFI) and temperature monitors Temperature is a major factor that determines the rate of spoilage of food and the rate at which pathogenic bacteria will grow in a food. Temperature potentiates the effect of other factors and, in many situations, is the factor most likely to fluctuate. Integrating temperature/time history, when combined with a thermal death model such as that proposed by Bigelow (1921) and Esty and Meyer (1922), has been standard practice to evaluate the efficacy of thermal processes in the canned food industry for many years. TFI has also been widely used to evaluate the hygienic equivalence of processes, particularly in the meat industry, and to estimate the safety and stability of foods during transit and storage [see McMeekin (2007) for a succinct review of the contributions of Dr. C.O. Gill in this field]. For these purposes, the choice of temperature monitoring device has been among chemical and physical monitors and electronic temperature loggers, with the latter becoming more widely used (Labuza, 2006). However, a biological temperature indicator was suggested as a potential solution to the non-alignment of chemical and biological reaction kinetics by McMeekin and Ross (1996). The innovative technology to translate the idea into a unique, practical and marketable device was provided by Cryolog with their TRACEO® Retail and (eO)® Take away products (www.cryolog.com). Development of the underpinning models was described by Ellouze et al. (2008).

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9

5

Regardless of the type of device chosen for temperature monitoring and integration, application is based on two innovative concepts, specified spoilage levels and relative rates, developed by Professor June Olley and her colleagues in the 1970s. The term “specified spoilage level”, which represents the time to reach a certain point in a sequence of events, was introduced by Olley and Ratkowsky (1973a,b). In the same publications, they refined a linear relative spoilage rate proposed by Nixon (1971) to one described by Arrhenius kinetics. Of note is mention of a “universal” spoilage model as it described a variety of deteriorative processes with similar activation energies, e.g. spoilage of various species of shark and mortality of abalone (Olley, 1971). Professors Olley and Ratkowsky have continued their search for a universal spoilage model since that time, most recently publishing a thermodynamically based model describing temperature effects on the growth rate of a wide range of bacteria (Ratkowsky et al., 2005). The mechanistic explanation is based on the denaturation of globular proteins at low and high temperatures. Subsequent, as yet unpublished, work has demonstrated the model also describes the influence of temperature on the growth of a wide range of Archaea.

Changing the workplace culture has led to significant success in reducing crashes in the aviation industry (Maurimo et al., 1995) and in surgical practices resulting in reduced traumatic outcomes for patients. A detailed treatise of human error was published by Reason (1990). There can be little argument that consumers should also be educated (or taught by rote) the basic elements of food safety and, further, that this should start early in life. In Australia, the Food Safety Information Council has this mission and estimates that ~25% of foodborne illness is attributable to consumers (www.foodsafety.asn.au). The first text on human error in food safety, authored by Frank Yiannas will be published in 2008 (Yiannas, in press). However, it appears that, regardless of significant efforts to educate consumers about food safety, the simple messages are not retained or, perhaps, are not prioritised when time-pressure is a significant factor of everyday life. Real-time predictive microbiology technologies, such as Cryolog, designed for application on individual packages at the retail and consumer levels of the supply chain, provide instant information, instant recommendations and instant decisions for time-pressured consumers.

4.3. Real-time reporting

5. Value analysis of predictive modelling R&D: the Refrigeration Index (RI) case study

The above describes the use of temperature loggers for process control in situ, but, during transport and storage, there may be problems with data recovery, retrospective analysis of information and manual examination for “progress reports”. However, these can be resolved by innovative technologies enabling real-time reporting. In Australia, Smart-Trace provides a good example of an electronic system and was described in detail by McMeekin et al. (2006b) and McMeekin (2007). The Cryolog temperature integrators (Section 4.2), which are based on the growth rate response of lactic acid bacteria to temperature fluctuations, also contain the important element of real-time reporting, in this case at the retail and consumer levels. The TRACEO® Retail product is an adhesive label applied on the barcode which is readable before the specific use-by-date of the product is reached and changes colour and becomes opaque, preventing scanning of the barcode, when the product has reached its use-by-date, or because the product has undergone a series of cold chain disruptions. The (eO)® Take away product is designed to operate on single units of refrigerated takeaway foods with a colour switch from green to red indicating end of shelf life. A decision to remove stock from display would be taken by retail staff, while a decision not to purchase or subsequently not to consume the product, would be taken directly by the consumer. Characteristic of the Cryolog products is that the colour transition from “bon” to “pas bon” occurs rapidly and within a few hours of the predicted end of shelf life for a specific product. This differentiates the Cryolog products from several other strip integrators in which colour changes gradually and interpretation of the extent of change is required (I-point AB Sweden, now Vitsab). The I-point indicator range is listed in Olley and Lisac (1985) and type 3270 was shown to be the most useful in the 0–15 °C range. Ironically, this type was not produced subsequently by Vitsab (June Olley, pers. comm.). However, it is important to understand that Cryolog products are temperature integrators which have taken into account the impact of the temperature fluctuations during distribution and storage on the quality of the product. This is unlike go/no go indicators that indicate a single deviation beyond a critical temperature or after a best before date calculated on the basis of storage time at a specified temperature. The decision to reduce the level of interpretation required by the consumer on the end of shelf life of a takeaway product (currently the purpose of one Cryolog product) has some interesting implications for making decisions on food safety. Whilst science and technology have been the traditional pillars of food safety, this needs to be combined with innovative approaches to educate and train industry staff and regulators to change the food safety culture and minimise human error (McMeekin, 2007).

The Refrigeration Index (RI) became the major outcome of significant research funding of predictive microbiology over several years by Meat and Livestock Australia (MLA) when it was incorporated into the revised Export Control (Meat and Meat Products) Order in 2005 by the Australian Quarantine and Inspection Service (AQIS, 2005) (see http://www.daffa. gov.au/aqis/export/meat/elmer-3). As a result, it is mandatory that the chilling of meat carcasses in Australian export abattoirs is based on an Escherichia coli growth model used to interpret cooling profiles measured by electronic temperature loggers which estimate the potential for E. coli growth and express it as the RI (www.mla.com.au). A value analysis of the return on research and development in predictive microbiology funded by MLA was carried out by the Centre for International Economics (www.thecie.com.au), taking into account investment and other inputs, outputs (measures of scientific worth), outcomes (applications arising from the R&D) and impact (return of investment for stakeholders) (see http://www.ml.com.au/HeaderAndFooter/AboutMLA/Corporate%2Bdocuments/Evaluation/default. htm#past%20summaries). The impacts for the raw meat sector of the Australian meat export industry over 30 years from adoption were: 1. Inputs: MLA AU$3.2 M, Australian Research Council and Australian Quarantine Inspection Service AU$0.6 M. 2. Outputs: Tool developed and validated by MLA; 11 scientific articles; extensive industry training by MINTRAC (Meat Industry Training Advisory Council). 3. Outcomes: RI mandated in the Export Control (Meat and Meat Products) Orders 2005. 4. Impacts: Lower compliance costs; increased regulatory confidence; more than 800 meat industry personnel trained. 5. Benefits: AU$44 M in red meat industry added value; benefit:cost ratio of 11:1; NAU$60 M in benefits to Australian consumers by 2028. The impacts for the processed meat industry over 30 years from adoption were: 1. Inputs: MLA partnership with the University of Tasmania as a preferred research provider; advice to the Meat Standards Committee (at that time the domestic market food safety regulator). 2. Outputs: RI tool and other software products; 11 refereed scientific papers; improved relationship between industry and regulators. 3. Outcomes: adoption of predictive microbiology to ensure efficient regulation and implementation of Australian standards. 4. Impacts: lower compliance costs; increased regulatory confidence; reduced illness and death from listeriosis.

6

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9

5. Benefits: AU$162 M in industry added value (pork industry major beneficiary); AU$281 M in social benefits to Australian consumers by 2028. In addition, the Australian meat export industry will gain considerable benefit from highly qualified regulatory and quality assurance staff trained to implement the new approach. The staff will be able to validate food safety programs and will have access to the latest relevant R&D outcomes from Australia and overseas. It is anticipated that this human resource will contribute to market access, particularly when involved with alternative procedures in an outcomes based regulatory environment. 6. Modelling food and other ecosystems Predictive food microbiology in its “modern” form has evolved over the last 30–35 years to the point where it was described (albeit too early, at that time) as a new paradigm in food microbiology (Labuza, 1994) and has been accepted by industry (Membré and Lambert, 2008) and regulatory authorities (AQIS, 2005) as an alternative, perhaps preferable, approach to predict the shelf life and safety of foods. However, there are other groups of microbiologists who are attempting to model different aquatic and terrestrial ecosystems to predict the microbial contribution to environmental degradation (and, conversely, develop bioremediation strategies), occurrence of algal blooms (Recknagel et al., 1997; Schoemann et al., 2005), Vibrio cholerae epidemics (Lobitz et al., 2000) and microbial roles in biogeochemical cycles which may impact on the production or feedback inhibition of greenhouse gases and, thereby, the acceleration or slowing of global warming and its consequences (Rose et al., 2001; Rosenberg and BenHaim, 2002; Bally and Garrabout, 2007). Unfortunately, for the discipline of predictive modelling as a whole, different groups of microbiologists tend to work within the confines of sub-discipline interest groups, in selection of journals for publication of their research, prefer to attend specialist conferences and select specific sessions at conferences with a wider range of topics on the program. Crystal ball-2007 [Environmental Microbiology 9(1): 1–11, 2007] featured leading researchers in the field of environmental microbiology who speculated on the technical and conceptual developments that will drive innovative research and open new vistas over the next few years. The scene was set by Curtis (2007) who wrote, “This field is hidebound by the difficulty of experimentation and is, therefore, contaminated by selfcongratulatory mathematical castles in the air with invented parameters and little verification”. And Hugenholtz (2007) observed that, to make sense of massive data sets, “modelling will assume a central role in microbial ecology. As a result, it will transition from a mainly qualitative, descriptive discipline to a quantitative predictive one”. As food microbiologists, do we require a modicum of self-examination to ensure we have not built castles, windmills or other structures without adequate foundations with which others may joust successfully? Have we invented the numbers and engaged only sparsely in validation? Alternatively, are the environmental writers not familiar with the voluminous literature in predictive food microbiology? Or have we, as predictive food microbiologists, been too insular and self-contented to spread the word? Clearly, these are questions worthy of informed and considered debate between food and environmental microbial modellers. 6.1. Food ecosystems The ability to model microbial behaviour in foods with reasonable accuracy may, in large part, be due to the characteristics of many food environments. These range from typical fresh foods which contain a high level of nutrients supporting rapid microbial growth rates and rapid lag phase resolution. They are also predominantly batch systems of short duration in which a dominant population of low diversity is selected. Such features greatly reduce variability in microbial response

times and uncertainty is minimised as the physico-chemical environment of many fresh foods is very well characterised. This represents the “classical” microbiology space in the dichotomy described by Bridson and Gould (2000) which is amenable to description by kinetic (deterministic) models. Fresh foods, because of rapid colonisation by microorganisms, have a short shelf life and over eons of time humans have devised ways to extend the keeping time and perhaps, by the same strategies, maximise food safety. A consequence of food preservation is that processed foods represent a harsher environment and the growth rate of the original population is reduced and the lag phase increased. A corollary is that the changed conditions select for a new spoilage association or different pathogen. The selection pressure continues as the conditions become increasingly harsh and eventually only organisms with a highly specialised physiology survive and grow. Examples include extreme halophiles in heavily salted fish (Chandler and McMeekin, 1989) and thermophiles on dairy processing equipment where temperatures N70 °C are attained in pasteurisers and evaporators (Langeveld et al., 1995). Only severe treatment with heat or ionising radiation can ensure sterility of a product, but at the expense of decreased sensory and nutritional properties. Any less than a complete kill will allow survivors, capable of repair and growth, to colonise the ecosystem again (Knight et al., 2004). Harsher environments and inevitably increased response times perhaps suggest a transition from the classical to the “quantal microbiology” space (Bridson and Gould, 2000). Note that quantal, not quantum (Graeme Gould, pers. comm), was used as no evidence has been presented to date that cannot be described by conventional physics (Bothma et al., 2007). What is certain, from a modelling viewpoint, is that as response time increases so does variability, eventually to the point where kinetic models are replaced by probability models (Ratkowsky et al., 1996). The point at which we choose to change approach from kinetic to probabilistic depends on the required level of confidence (i.e. probability of “failure” and the consequences of failure (i.e. the “risk”). 6.2. Ecosystems other than food Environmental microbial ecologists are often confronted with describing and/or quantifying microbial roles and activity in ecosystems that are complex and exist on many scales from microcosms to oceans. In contrast to most food environments they are open (continuous flow) systems rather than the batch systems most frequently encountered in food microbiology. Also, as opposed to food ecosystems, they are frequently nutrient depleted and, thus, support low populations but with diverse climax communities that employ various feedback (homeostatic) mechanisms to maintain equilibrium in the community (Atlas and Bartha, 1987). Whilst food ecosystems come and go, in a time frame of days, weeks or months, many environmental systems exist over very long time frames, both in terrestrial and aquatic habits (Whitman et al., 1998; Karner et al., 2001). Environmental microbiologists also appear to be much more concerned by the implications of unculturable microbiota and microbial diversity than food microbiologists (Prosser et al., 2007). Many natural microbial environments are subject to regular fluctuations which determine the growth or death of components of the ecosystems. An example is the seasonal appearance of Vibrio parahaemolyticus in seawater in summer, in response to rising temperatures, resulting in an increase in foodborne illness due to this organism (CDC, 1998; Cook et al., 2002). On the decline side of the ledger, E. coli has been shown to be inactivated much more rapidly under light than under dark conditions due to the effect of UV radiation with wavelengths up to 450 nm contributing to inactivation. This was demonstrated by several groups ~ 1980 (Chamberlin and Mitchell, 1978; Fujioka et al., 1981; McCambridge and McMeekin, 1981), but it is salutary to note that the same phenomenon, same effective depths and comparable rates were reviewed in the book

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9

“Microorganisms in Water” by Frankland and Frankland (1894) in which Chapter IX dealt with the “Action of Sunlight on Microoganisms in Water”. Whilst preventing foodborne illness and food spoilage is socially responsible and saves multi-million dollars for consumers and the food industry, environmental microbial ecology is central to understanding microbial roles in biogeochemical cycles. These include those releasing or sequestering gases associated with “big picture” science questions such as, is global warming due to anthropogenic activity or natural variation in climatic cycles? Because of worldwide concern, current levels of research interest and, therefore, funding, will place environmental microbial ecologists in pole position to advance fundamental understanding of microbial ecology questions. However, it will be interesting to discover if predictive food microbiology can provide direction to future studies in environmental microbial ecology, should ever the twain meet. 7. Summary and general projections The research pioneers of predictive food microbiology, including Tino Genigeorgis, Terry Roberts, Bob Buchanan, June Olley and David Ratkowsky, like very troublesome microorganisms, showed remarkable persistence in developing the field. This was achieved in the face of much scepticism that accurate models could be constructed, and used subsequently, to predict microbial population behaviour by considering only environmental factors. Indeed, one leading food microbiologist went so far as to describe predictive modelling as numerology. While the title Predictive Microbiology may conjure up images of crystal-balls, it certainly is not the study of the supposed occult significance of numbers. In all probability, such scepticism encouraged the early researchers to continue and to build networks of microbiologists, statisticians, process engineers, food technologists, electronics experts, et al., required to provide the strategic science and innovative technologies needed for the concept (Scott, 1937) to succeed. At this point in time, it is reasonable to conclude that we have succeeded with widespread acceptance of predictive models which, in some jurisdictions, are now incorporated in food safety legislation and with major food manufacturers using boundary models to formulate mildly preserved products with required stability, i.e. a quantified hurdle concept. Predictive models and their attendant databases have also been crucial in the operation of quantitative microbial risk assessment which is likely to have “drowned” in uncertainty in the absence of the knowledge summarised in models. The quantitative approach to microbial ecology has also identified areas of particular interest for detailed physiological and molecular studies and these links will inevitably connect the modellers with systems biologists and network analysts who will be required to interpret the data deluge. However, most of this research has been carried out in the predictive food microbiology “club”, which evolved from an informal group of scientists with common research interests who met for the first time at an international conference in Tampa, Florida in 1992 and for the fifth time in Athens, Greece in 2007. The group in 2008 was formalised as a Professional Development Group of the International Association of Food Protection. Now that we have reached a state of maturity and acceptance in the food microbiology community, the time is right to expand our horizons and to interact with microbiologists developing models to describe the microbial role in ecosystems other than food. Such interactions will expose us to different thought processes, to new experimental difficulties, to spatial and temporal boundaries currently way beyond our “ken”. Nevertheless, these interactions and the challenges posed will benefit both food microbiology and environmental microbiology. Provided, of course, that we maintain commitment to strategic research as a basis to develop theory and mechanisms and couple this with innovative technologies to meet the great expectations that will arise from discovering and understanding more of the microbial world.

7

The preceding sentences express a general view about the benefits of networking and collaboration. But general views and the objectives of high level vision or mission statements often dissipate due to lack of a specific task or case study to demonstrate the probability of a good outcome. Thus, we choose to conclude by suggesting a specific research direction. 8. Moving forward through improved models of microbial persistence and recovery 8.1. A common research interest An area of common interest in food and environmental microbiology is the enhanced ability of some cells to persist under very harsh conditions. Spores are the ultimate survival machines resisting severe physical and chemical challenges by dint of their tank-like construction. However, vegetative cell populations of E. coli have been demonstrated to contain a subpopulation of resistant cells in a genetically homogeneous population (Balaban et al., 2004); Smits et al. (2006) identified phenotypic variation and the role of feedback regulation leading to states described as multistationarity and multistability; and toxin– antitoxin systems, referred to as suicide modules, are chromosomallyencoded genes that can mediate self-destruction of a cell (Aizenman et al., 1996; Engelberg-Kulka and Glaser, 1999; Gerdes, 2000). The hypothesis is that these systems in bacterial cells are analogous to the apoptotic machinery in multicellular organisms and that sacrificing individual cells benefits the population as a whole. This serves to emphasise that survival strategies operate at both single cell and population levels and suggests further study of “multicellularity” in bacterial populations (Shapiro, 1998). At the population level, biofilms exemplify a ‘strength in numbers’ tactic with the protection inherent in individual cells supplemented by extracellular defences. Within these maze-like structures, akin to fortified mediaeval villages with myriads of underground passages, the microbial inhabitants communicate, extract nutrients, may cometabolise and effectively resist the ingress of inimical substances. The study of persistence and biofilms will also require different types of modelling, e.g. attachment to and removal from surfaces (Schaffner, 2003) and for the “-omics” researchers, differentiation of persistent cells from others in the population would be a worthwhile research objective. And, while dealing with persistence, the mechanisms of recovery from various “dormant” states will present significant research challenges. Optimising recovery on various media has been a long-term interest in food microbiology, but many studies have been conducted empirically rather than progressing towards a mechanistic explanation such as the addition of catalase or pyruvate to scavenge free radicals (Mackey and Derrick, 1982; Mackey and Seymour, 1987). 8.2. Selecting organisms for collaborative studies Are there organisms that would be suitable for parallel study by environmental and food microbiologists? From the viewpoint of food safety, two of the most troublesome organisms in food microbiology are toxigenic E. coli and Listeria monocytogenes. The latter is well known as an environmental organism and has been isolated from soil, vegetation, sewage and water. From these repositories they may colonise foodprocessing environments with some strains displaying remarkable persistence (Holah et al., 2004). In the same publication, colonisation of food processing premises by persistent strains of E. coli was demonstrated and, subsequently, the question “Escherichia coli 0157: burger bug or environmental pathogen?” was posed by Strachan et al. (2006) who concluded that acquisition of infection in NE Scotland was 100-times more likely to occur when visiting a pasture than eating a burger. The pathogenic marine vibrios, V. cholerae, V. parahaemolyticus and V. vulnificus, are another microbial group in which shared research interests between environmental and predictive food microbiologists

8

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9

could readily occur. Recently, FAO/WHO reported the results of microbial risk assessment on these organisms (V. cholerae in shrimp in international trade; V. parahaemolyticus in seafood and V. vulnificus in raw oysters) (FAO/WHO, 2006). As a result, much information was assembled and important data gaps were identified. For example, the risk assessment relied on a single report of the effect of temperature on the growth rate of V. parahaemolyticus in bacteriological broth (Miles et al., 1997). This was adjusted by a factor of four for consistency with a single report showing a slower growth rate for V. parahaemolyticus at 26 °C in oysters (Gooch et al., 2002). Thus, the need is emphasised for robust models and databases that can fill major data gaps, reduce uncertainty and increase confidence in the estimated levels of risk. The seasonal occurrence of marine vibrios, their widespread occurrence in marine environments, the potential for their temporal and spatial distribution to be affected by increased water temperatures and their association with foodborne illnesses ranging from those associated high-risk individuals, as with V. vulnificus, to pandemics with V. cholerae 01 and 0139, make these organisms a suitable group for collaborative modelling studies. For example, Lobitz et al. (2000) reported the use of satellite data to monitor temporal and spatial changes in cholera cases in Bangladesh from 1992–1995, reporting that sea surface temperature, measured by satellite infrared measurement, correlated with cholera case data. Remote sensing technology was also used by Phillips et al. (2007) to evaluate risk associated with V. parahaemolyticus in Gulf Coast oysters. 8.3. Models of persistence and recovery to inform the broader horizon of microbial ecology Persistence and recovery are also key features in microbial dispersal and colonisation of new habitats, giving rise to important questions in microbial ecology such as: - speciation: are species ubiquitous because of unrestricted dispersal or are they restricted to island communities because of barriers to dispersal; are there ecospecies (ecotypes) as proposed by Cohan (2002). - diversity: what is the extent of microbial diversity, how many species cannot be cultivated and what is their role in natural environments? (Prosser et al., 2007). - “resting stages”: what is the significance of inactive cells for survival of a species: Is there such a state as viable but non culturable (VBNC), or is recovery of a small number of very persistent cells responsible for the reappearance of viable cells in an ecosystem? Should the term be ‘active but non culturable’ (ABNC) as proposed by Kell et al. (1998) as an operational definition and to remove the oxymoron status of VBNC? [For more detail on the points above, and other pertinent questions, see Prosser et al. (2007)]. Vibrios and related organisms have featured prominently in the A or VBNC debate, going back to the research of Morita and colleagues working with the Vibrio-like organism ANT 300. Now classified as Moritella marina, ANT 300 gained “celebrity status” amongst marine microbial ecologists with its ability, following starvation, to form dwarf cells (~11-fold reduction in volume), ditch all reserve and capsular material and low molecular weight carbohydrates and decrease endogenous respiration by 99% (Novitsky and Morita, 1976). Conversely, the dwarf cells multiplied rapidly, produced a sheathed flagellum and displayed a wide range of chemotactic responses and remarkable substrate capture ability (Torrella and Morita, 1981). This appears to contradict the self-preservation and nutritional competence (SPANC) balance which holds that, in E. coli, high stress resistance is associated with reduced ability to compete for growth substrates at suboptimal concentrations (Ferenici and Spira, 2007). So, perhaps appropriately, we end with a conundrum, further emphasising that much remains to be understood in both environmental and food microbial ecology. Perhaps

the ‘hybrid vigour’ obtained by combining different lineages of theory and practice will yield increased insight and innovation to define the crucial role of microbial activity in the biosphere? References Aizenman, E., Engelberg-Kulka, H., Glaser, G., 1996. An Escherichia coli chromosomal “addiction module” regulated by 3′,5′-bispyrophosphate: a model for programmed bacterial cell death. Proceedings of the National Academy of Science USA. 93, 6059–6063. AQIS (Australian Quarantine and Inspection Service) 2005. Export Control (Meat and Meat Products) Orders 2005 (http://www.daffa.gov.au/aqis/export/meat/elmer-3). Atlas, R.M., Bartha, R., 1987. Microbial Ecology: Fundamentals and Applications, 2nd Edition. Benjamin/Cummins Publishing Company Ltd., Menlo, California, p. 533. Balaban, N.Q., Merrin, J., Chait, R., Kowalik, L., Leibler, S., 2004. Bacterial persistence as a phenotypic switch. Science 305, 1622–1625. Bally, M., Garrabout, J., 2007. Thermodependent bacterial pathogens and mass mortalities in temperate benthic communities: a new case of emerging disease linked to climate change. Global Changes in Biology 13, 2077–2088. Barabasi, A.-L., 2002. Linked: How Everything is Connected to Everything Else and What it Means. Penguin Books Ltd., London, UK, 304 p. Bigelow, W.D., 1921. The logarithmic nature of thermal death time curves. Journal of Infectious Diseases 29, 528–536. Bongaarts, J., 2005. Five period measures of longevity. Demographic Research 13, 547–558. Bothma, J., Gilmore, J., McKenzie, R.H., 2007. Modelling quantum decoherence in biomolecules. Quantum Aspects of Life. World Scientific Book, Chapter 1, pp. 5–29. Bridson, E.Y., Gould, G.W., 2000. Quantal microbiology. Letters in Applied Microbiology 30, 95–98. Centers for Disease Control and Prevention (CDC), 1998. Outbreak of Vibrio parahaemolyticus infections associated with eating raw oysters—Pacific Northwest 1997. Morbidity and Mortality Weekly Report 47, 457–462. Chamberlin, C.E., Mitchell, R., 1978. A decay model for enteric bacteria in natural waters. In: Mitchell, R. (Ed.), Water Pollution Microbiology, Vol. 2. Wiley, New York, pp. 325–348. Chandler, R.E., McMeekin, T.A., 1989. Combined effect of temperature and salt concentration/water activity on the growth rate of Halobacterium spp. Journal of Applied Bacteriology 67, 71–76. Cohan, F.M., 2002. What are bacterial species? Annual Reviews in Microbiology 56, 457–487. Cook, D.W., O’Leary, P., Hunsucker, J.C., Sloan, E.M., Bowers, J.C., Blodgett, R.J., DePaola, A., 2002. Vibrio vulnificus and Vibrio parahaemolyticus in U.S. retain shell oysters: a national survey, June 1998 to July 1999. Journal of Food Protection 65, 79–87. Curtis, T., 2007. Theory and the microbial world. Environmental Microbiology 9, 1. Ellouze, M., Pichaud, M., Bonaiti, C., Coroller, L., Couvert, O., Thuault, D., Vaillant, R., 2008. Modelling pH evolution and lactic acid production in the growth medium of a lactic acid bacterium. Application to set a biological TTI. International Journal of Food Microbiology (submitted to this Special Issue). Engelberg-Kulka, H., Glaser, G., 1999. Addiction modules and programmed cell death and antideath in bacterial cultures. Annual Review of Microbiology 53, 43–70. Esty, J.R., Meyer, K.F., 1922. The heat resistance of spores of B. botulinus and related anaerobes. Journal of Infectious Diseases 31, 650–663. Ferenici, T., Spira, B., 2007. Variation in stress responses within a bacterial species and the indirect costs of stress resistance. Annals of the New Your Academy of Science 1113, 105–113. Food and Agriculture Organisation/World Health Organisation 2006. Risk assessment of Vibrio parahaemolyticus in seafood: interpretative summary and technical report (www.who.int/foodsafety/publication/micro/en/index.html). Frankland, P., Frankland, P., 1894. Microorganisms in Water; Their Significance, Identification and Removal. Chapter IX. Action of Sunlight on Microorganisms in Water. Longmans Green and Co., London. Frederiksen, M., Osterberg, C., Silberg, S., Larsan, E., Bremner, A., 2002. Info-Fisk. Development and validation of an internet based traceability system in a Danish domestic fresh fish chain. Journal of Aquatic Food Product Technology 11, 13–34. Fujioka, R.S., Hashimoto, H.H., Siwak, E.B., Young, R.F.P., 1981. Effect of sunlight on survival of indicator bacteria in seawater. Applied and Environmental Microbiology 41, 690–696. Geeraerd, A.H., Valdramidis, V.P., Devlieghere, F., Bernaerts, H., Debevere, J., Van Impe, J.F., 2004. Development of a novel approach for secondary modelling in predictive microbiology: incorporation of microbiological knowledge in black box polynomial modelling. International Journal of Food Microbiology 91, 229–244. Gerdes, K., 2000. Toxin–antitoxin modules may regulate synthesis of macromolecules during nutritional stress. Journal of Bacteriology 182, 561–572. Gompertz, B., 1825. On the nature of the function expressive of the law of human mortality and on a new mode of determining the value of life contingencies. Philosophical Transactions of the Royal Society, London 115, 513–585. Gooch, J., De Paola, A., Bowers, J., Marshall, D., 2002. Growth and survival of Vibrio parahaemolyticus in postharvest American oysters. Journal of Food Protection 65, 970–974. Gross, A.J., Clark, V.A., 1975. Survival Distributions: Reliability Applications in the Biomedical Sciences. Wiley, New York. Holah, J.T., Bird, J., Hall, K.E., 2004. The microbial ecology of high-risk, chilled food factories; evidence for persistent Listeria spp. and Escherichia coli strains. Journal of Applied Microbiology 97, 68–77. Hugenholtz, P., 2007. Riding giants. Environmental Microbiology 9, 5. Karner, M.B., de Long, E.F., Karl, D.M., 2001. Archaeal dominance in the mesopelagic zone of the Pacific Ocean. Nature 409, 5107–5109.

T. McMeekin et al. / International Journal of Food Microbiology 128 (2008) 2–9 Kell, D.B., Kaprelyants, A.S., Weichart, D.H., Harwood, C.R., Barer, M.R., 1998. Viability and activity in readily culturable bacteria: a review and discussion of the practical issues. Antonie van Leeuwenhock 73, 169–187. Knight, G.C., Nicol, R.S., McMeekin, T.A., 2004. Temperature step changes: a novel approach to control biofilms of Streptococcus thermophilus in a pilot plant-scale cheese-milk pasteurisation plant. International Journal of Food Microbiology 93, 305–318. Krist, K., Ross, T., McMeekin, T.A., 1998. Final optical density and growth rate: effects of temperature and NaCl differ from acidity. International Journal of Food Microbiology 43, 195–204. Labuza, T.P., 1994. Microbial modelling—a new paradigm in food science. Food Technology 48, 16. Labuza, T., 2006. Time-temperature integrators and the cold chain: what is next? In: Kreyenschmidt, J., Petersen, B. (Eds.), Proceedings of the 2nd International ColdChain Management Workshop, Universitasdruckeri, Bonn, pp. 71–78. Lambert, R.J.W., Bidlas, E., 2007. A study of the Gamma hypothesis: predictive modelling of the growth and inhibition of Enterbacter sakazakii. International Journal of Food Microbiology 115, 204–213. Langeveld, L.P.M., van Montfort-Quasig, R.M.G.E., Werkamp, A.H., Waal Jr., E.W.I., Wever, J.S., 1995. Adherence, growth and release of bacteria in a tube heat exchanger for milk. Netherlands Milk Dairy Journal 49, 207–220. Leistner, L., 2000. Basic aspects of food preservation by hurdle technology. International Journal of Food Microbiology 55, 181–186. Leistner, L., Gorris, L.G.M., 1995. Food preservation by hurdle technology. Trends in Food Science and Technology 6, 41–46. Lobitz, B., Beck, L., Huq, A., Wood, B., Fuchs, G., Faruque, A.S.G., Colwell, R., 2000. Climate and infectious disease: use of remote sensing for detection of Vibrio cholerae by indirect measurement. PNAS 97, 1438–1443. Mackey, B.M., Derrick, C.M., 1982. A comparison of solid and liquid media for measuring the sensitivity of heat-injured Salmonella typhimurium to selenite and tetrathiomate media and the time needed to recover resistance. Journal of Applied Bacteriology 53, 233–242. Mackey, B.M., Seymour, D.A., 1987. The effect of catalase on recovery of heat-injured DNA-repair mutants of Escherichia coli. Journal of General Microbiology 133, 233–242. Maurimo, D., Reason, J., Johnson, J., Lee, R.B., 1995. Beyond Aviation Human Factors. Ashgate Publishing, UK, p. 169. McCambridge, J., McMeekin, T.A., 1981. Effect of solar radiation and predacious microorganisms on survival of faecal and other bacteria. Applied and Environmental Microbiology 41, 1083–1087. McMeekin, T.A., 2007. Predictive Microbiology: quantitative science delivering quantifiable benefits to the meat industry and other food industries. Meat Science 77, 17–27. McMeekin, T.A., Ross, T., 1996. Shelf life prediction: status and future possibilities. International Journal of Food Microbiology 33, 65–83. McMeekin, T.A., Olley, J., Ross, T., Ratkowsky, D.A., 1993. Predictive Microbiology: Theory and Application. Research Studies Press and John Wiley and Sons, Taunton, UK, pp 340. McMeekin, T.A., Olley, J., Ratkowsky, D.A., Ross, T., 2002. Predictive microbiology: towards the interface and beyond. International Journal of Food Microbiology 73, 395–407. McMeekin, T.A., Szabo, L., Ross, T., 2005. Connecting science with technology to improve microbial food safety management. International Review of Food Science and Technology 125–130 Winter 2005/2006. McMeekin, T.A., Baranyi, J., Bowman, J.P., Dalgaard, P., Kirk, M., Ross, T., Schmid, S., Zwietering, M.H., 2006a. Information systems in food safety management. International Journal of Food Microbiology 112, 181–194. McMeekin, T.A., Smale, N., Jenson, I., Ross, T., Tanner, D.B., 2006b. Combining microbial growth models with near real-time temperature monitoring technologies to estimate the shelf-life and safety of foods during processing and distribution. In: Kreyenschmidt, J., Petersen, B. (Eds.), Proceedings of the 2nd International Coldchain Management Workshop, Universitasdruckeri, Bonn, pp. 43–52. Membré, J.M., Lambert, R.J.W., 2008. Predictive modeling techniques in industry: from food design up to risk assessment. International Journal of Food Microbiology (submitted to this Special Issue). Miles, D.W., Ross, T., Olley, J., McMeekin, T.A., 1997. Development and evaluation of a predictive model for the effect of temperature and water activity on the growth rate of Vibrio parahaemolyticus. International Journal of Food Microbiology 38, 133–142. Neimeyer-Stein, N., 2006. Labelling product packages with lost-cost secure optical identicodes. In: Kreyenschmidt, J., Petersen, B. (Eds.), Proceedings of the 2nd International Cold-Chain Management Workshop, Universitasdruckeri, Bonn, pp. 83–90.

9

Nixon, P.A., 1971. Temperature integration as a means of assessing storage conditions. Report on Quality in Fish Products. Seminar No. 3, Fishing Industry Board, New Zealand, pp. 33–44. Novitsky, J.A., Morita, R.Y., 1976. Morphological characterisation of small cells resulting from nutrient starvation of a psychrophilic marine vibrio. Applied and Environmental Microbiology 32, 617–622. Olley, J., 1971. Handling of abalone. Report on Quality of Fish Products. Seminar No. 3, Fishing Industry Board, New Zealand, pp. 89–95. Olley, J., Ratkowsky, D.A., 1973a. Temperature function integration and its importance in the distribution and storage of flesh foods above the freezing point. Food Technology Australia 25, 66–73. Olley, J., Ratkowsky, D.A., 1973b. The role of temperature function in monitoring fish spoilage. Food Technology New Zealand 8, 13–17. Olley, J., Lisac, H., 1985. Time/temperature monitors. Infofish Marketing Digest 3, 45–47. Oomes, S.J.C.M., va Zuijlen, A.C.M., Hehenhamp, J.O., Witsenboer, H., van der Vossen, J.M.B.M., Brul, S., 2007. The characterisation of Bacillus spores occurring in the manufacturing of (low acid) canned products. International Journal of Food Microbiology 120, 85–94. Phillips, A.M.B., DePaola, A., Bowers, J., Ladner, S., Grimes, D.J., 2007. An evaluation of the use of remotely sensed parameters for prediction of incidence and risk associated with Vibrio parahaemolyticus in Gulf Coast Oysters (Crassostrea virginica). Journal of Food Protection 70, 879–884. Prosser, J.I., Bohannan, B.J.M., Curtis, T.P., Ellis, E.J., Firestone, M.K., Freckleton, R.P., Green, J.L., Green, L.E., Killham, K., Lennon, J.J., Osborn, A.M., Solan, M., van der Gast, C.J., Young, J.P. W., 2007. The role of ecological theory in microbial ecology. Nature 5, 384–392. Ratkowsky, D.A., 1993. Principles of nonlinear regression modelling. Journal of Industrial Microbiology 12, 195–199. Ratkowsky, D.A., Ross, T., 1995. Modelling the bacterial growth-no growth interface. Letters in Applied Microbiology 20, 29–33. Ratkowsky, D.A., Ross, T., Macario, N., Dommett, T.W., Kamperman, L., 1996. Choosing probability distributions for modelling generation time variability. Journal of Applied Bacteriology 80, 131–137. Ratkowsky, D.A., Olley, J., Ross, T., 2005. Unifying temperature effects on the growth rate of bacteria and the stability of globular proteins. Journal of Theoretical Biology 233, 351–362. Reason, J., 1990. Human Error. Cambridge University Press, p. 302. Recknagel, F., French, M., Harkonen, P., Yabunaka, K., 1997. Artificial neural network approach for modelling and prediction of algal blooms. Ecological Modelling 96, 11–28. Rose, J.B., Epstein, P.R., Lipp, E.K., Sherman, B.H., Bernard, S.M., Patz, J.A., 2001. Climate variability and changes in the United States: potential impacts on water- and foodborne diseases caused by microbiologic agents. Environmental Health Perspectives 109, 211–221 Supplement 2. Rosenberg, E., Ben-Haim, Y., 2002. Microbial diseases of corals and global warming. Environmental Microbiology 4, 318–326. Schaffner, D.W., 2003. Models—what comes after the next generation? In: McKellar, R.C., Lu, X. (Eds.), Modelling Microbial Responses in Food. CRC Press Inc., Boca Raton, FL, pp. 303–312. Schoemann, V., Becquevort, S., Stefels, J., Rosseau, W., Lancelot, C., 2005. Phaeocystis blooms in the global ocean and their controlling mechanisms. Journal of Sea Research 52, 43–66. Scott, W.J., 1937. The growth of microorganisms on ox muscle. I. The influence of temperature. Journal of the Council of Scientific and Industrial Research, Australia 10, 338–350. Shapiro, J.A., 1998. Thinking about bacterial populations as multicellular organisms. Scientific American 258, 82–89. Smits, W.K., Kuipers, O.P., Veening, J.-W., 2006. Phenotypic variation in bacteria: the role of feedback regulation. Nature Reviews Microbiology 4, 259–271. Strachan, N.J.C., Dunn, G.M., Locking, M.E., Reid, T.M.S., Ogden, I.D., 2006. Escherichia coli 0157: burger bug or environmental pathogen? International Journal of Food Microbiology 112, 129–137. Teriokhin, A.T., Budilova, E.V., Thomas, F., Guegan, G.T., 2004. Worldwide variation in life-span sexual dimorphism and sex-specific mortality rates. Human Biology 76, 623–641. Torrella, F., Morita, R.Y., 1981. Microcultural study of bacterial size changes and microcolony and ultramicrocolony formation by heterotrophic bacteria in seawater. Applied and Environmental Microbiology 41, 518–527. Whitman, W.B., Coleman, D.C., Wiebe, W.J., 1998. Prokaryotes, the unseen majority. Proceedings of the National Academy of Science, USA 95, 6578–6583. Yiannas, F., (in press). Food Safety Culture: Creating a behaviour-based food safety management system, 70 pp in Food Microbiology and Food Safety, Springer Scientific.