Bioaugmentation for bioremediation: the challenge of strain selection


*E-mail; Tel. (+44) 0 1865 281630; Fax (+44) 0 1865 281696.


Despite its long-term use in bioremediation, bioaugmentation of contaminated sites with microbial cells continues to be a source of controversy within environmental microbiology. This largely results from its notoriously unreliable performance record. In this article, we argue that the unpredictable nature of the approach comes from the initial strain selection step. Up until now, this has been dictated by the search for catabolically competent microorganisms, with little or no consideration given to other essential features that are required to be functionally active and persistent in target habitats. We describe how technical advances in molecular biology and analytical chemistry, now enable assessments of the functional diversity and spatial distribution of microbial communities to be made in situ. These advances now enable microbial populations, targeted for exploitation, to be differentiated to the cell level, an advance that is bound to improve microbial selection and exploitation. We argue that this information-based approach is already proving to be more effective than the traditional ‘black-box’ approach of strain selection. The future perspectives and opportunities for improving selection of effective microbial strains for bioaugmentation are also discussed.


There are few issues in environmental biotechnology that generate more controversy than the use of bioaugmentation in bioremediation. Some see it as an expensive and ineffective equivalent to voodoo medicine, while others extol its virtues and even derive a living from its commercialization. The rationale underlying bioaugmentation in bioremediation is simple: the augmentation of catabolically-relevant organisms to hasten remediation. But despite its apparent simplicity there have been many failures, and these have been well detailed and reviewed (Goldstein et al., 1985; Stephenson and Stephenson, 1992; Bouchez et al., 2000; Vogel and Walter, 2001; Wagner-Döbler, 2003). Often, failures have been simply due to choosing bioaugmentation as the remediation approach in the first place. In these cases the absence of relevant catabolic genes within the indigenous microbial community is not the factor limiting contaminant degradation. Biodegradation can be inhibited by a plethora of factors, such as pH and redox, the presence of toxic co-contaminants, concentration, or the absence of key co-substrates, to name but a few (Barriault and Sylvestre, 1993; Harms and Zehnder, 1995). The purpose of this article is not to review the many instances where bioaugmentation failures could have been predicted by undertaking microbial and chemical analyses to identify possible constraints to degradation. Our focus is on those more challenging cases where, after appropriate analysis, there are no obvious constraints to microbial degradation, yet bioaugmentation still fails. We believe that the route of these failures is summarized by Vogel and Walter (2001), ‘Although microbial ecology issues are among the most important in bioaugmentation approaches, unfortunately, they are rarely addressed’. The authors argue that, as with indigenous populations, a broad range of environmental parameters, together with the phenotypic characteristics of the strains and procedures for introduction, culminate to determine the activity, persistence and performance of bioaugmented strains. It is the lack of consideration of such determinants and the complete and sole preoccupation with the search for vigorous degradative strains that is the root of many failures.

In this article, we argue that ecological considerations should be a priority, even before the application of inocula: that is at the strain selection stage. Vital to these ecological considerations is the relative spatial and temporal abundance of potential source populations and their ability to tolerate the prevailing conditions in target habitats. Until recently, initial strain selection has been based on a single key criterion, degradation ability, with no consideration of the potential ability of strains to proliferate and be active in sites to which they were applied. We propose that strains selected from populations that have low abundance, or are transient within the source or target habitats, are less likely to succeed as inocula, when compared with those that are spatially and temporally ubiquitous. We discuss the challenge of selecting strains for reliable exploitation in bioaugmentation applications: a key stage, but sparsely dealt with in most relevant papers. We briefly summarize some of the more recent technical advances that now enable strain selection to be based on an in situ understanding of organism abundance, functional activity and population dynamics in the habitat from which they are derived. We go on to provide evidence as to why a more informed approach is already proving to be more effective, than the traditional ‘black-box’ attitude to selecting ‘good strains’. The future opportunities these technological advances provide for selecting well-adapted and catabolically  competent  strains,  for  degradation per  se, as well as their potential as vehicles for delivering and maintaining degradative genes in the target habitat are discussed.

The challenge

Successful application of bioaugmentation techniques is dependent on the identification and isolation of appropriate microbial strains, and their subsequent survival and activity, once released into the target habitat. Sourcing microbial strains for bioaugmentation has typically been achieved, for at least the past 100 years, by selective enrichment (Beijerinck, 1901). This involves strains from polluted samples being enriched to grow, relative to the background community, in culture, using the target contaminant as the sole enriching carbon or nitrogen source. The procedure results in the selection of strains that express the required degradation ability: in the specific conditions of the enrichment culture, at least. However, such enriched populations are not necessarily typical or representative of indigenous communities in the target habitat and could equally, by chance, be derived from transient populations. It seems likely that this latter point is the crux of many bioaugmentation failures. The problem is that the enrichment procedure is unlikely to have any influence on other traits that are also required for strains to be competitive and effective in the target environment. These additional traits are required to survive prevailing, often fluctuating, environmental conditions (e.g. moisture, nutrients, redox, pH and osmotic factors) and, of course, competition from indigenous microbial populations, among many others stresses (Recorbet et al., 1992; England et al., 1993). We believe it is this initial selection step, and specifically poor knowledge of the composition and population dynamics of source communities, as well as those in the target habitats (competitors in waiting), that underlie many inocula failures.

An alternative approach, which would probably increase success rates, in terms of persistence and activity, would be to base the initial selection step on a priori knowledge of the ubiquity, population dynamics and spatial and temporal distribution of microbial communities in sampled habitats. It seems logical that a strain, derived from a population that is temporally and spatially prevalent in a specific type of habitat, is more likely to persist as an inoculum when reintroduced, than one that is transient or even alien to such a habitat. Once abundant populations have been identified, the second phase of the selection procedure should then be to identify strains which can degrade the target contaminant. Indeed, there is evidence to suggest that such a strategy would aid inocula survival. Belotte and colleagues (2003) isolated bacteria from a range of locations in a forest soil. When vigorous growers were cultured and reintroduced into the environment, they tended to colonize the sites from which they originated, providing strong evidence for local adaptation. The authors estimated that isolates were 50% fitter when reintroduced into their home sites, and that fitness diminished exponentially with distance from their origin (Belotte et al., 2003).

We are not advocating that the selective enrichment method should be made redundant. The approach is essential in order to derive ‘superior strains’ and so will continue to be an essential step for obtaining exploitable strains (Singer et al., 2005a). Nor are we suggesting that intensive microbial community analysis be taken for every site targeted for bioaugmenation. Tailoring inocula for every site and application would just not be practicable. However, we are suggesting that selection of strains should be taken on the basis of some understanding of the kind of microbial communities present in the source habitat, and preferably with some knowledge of the type of organisms that are common in the target habitat. With the introduction of superior methods for undertaking more thorough assessments of microbial communities, this knowledge-based approach is becoming increasingly more feasible.

Identification of key populations and genes in situ

In recent years technological advances in molecular microbial ecology and analytical chemistry have been brought together, enabling us to identify in situ populations, and even individual cells, responsible for undertaking specific processes. Using molecular approaches, such as denaturing gradient gel electrophoresis (DGGE) (Muyzer et al., 1993), terminal restriction fragment length polymorphism (tRFLP) (Liu et al., 1997) and length-heterogeneity polymerase chain reaction (LH-PCR) (Suzuki et al., 1998), we are now in a better position to obtain a more comprehensive assessment of the composition and structure of microbial communities in the environment. In addition, other complementary techniques can provide more direct assessments of microbial community composition. For example, fluorescent in situ hybridization (FISH) (Delong et al., 1989) uses a fluorochrome-labelled oligonucleotide to identify specific populations in situ, and obtain direct measures of their relative abundance. More recently stable isotope probing (SIP) (Boschker et al., 1998; Radajewski et al., 2000) has emerged as a potentially powerful tool, which employs stable isotopes such as 13C, to determine exactly which organisms assimilate specific contaminants. Extraction of labelled fatty acids, DNA and/or RNA, allows accurate identification and differentiation of the active members of the community. With further development this should become the most powerful of the new techniques discussed; although as yet, it does not provide a reliable estimate of the relative abundance of the catabolically active forms (Manefield et al., 2002). However, SIP has already been applied in the field, where it successfully identified the populations that drive the remedial process (Jeon et al., 2003). In an exciting recent development, this approach has been combined with Raman confocal microscopy, lending insight into those populations responsible for contaminant catabolism to be differentiated, at the individual cell level (Huang et al., 2004; Singer et al., 2005b).

Better selection gives better performance

The techniques described, in particularly SIP, are new and not yet routinely and widely used to scrutinize communities, so it is really too early to predict how they might revolutionize the selection, detection and exploitation of strains for bioaugmentation. However, several studies are provided as evidence of how systematic and careful selection process for competitive bacterial strains might be used to improve bioaugmentation. For instance, in an extensive field study 690 characterized isolates of fluorescent pseudomonads from a single site were analysed to determine the genetic composition using RFLP rRNA (ribotyping) analysis, and dynamics of the population over several seasons (Goddard et al., 2001). The population was found to be highly heterogeneous: the 690 characterized isolates consisted of 385 ribotypes, and notably most were transient, being detected only once. Approximately 26 ribotypes were detected more frequently (spatially and temporally); one isolate, ribotype A, was ubiquitous in the samples analysed. Ribotype A was subsequently genetically tagged (lacZY with kanamycin resistance), inoculated in lab-based field soil and found to persist significantly better than randomly selected (transient) ribotypes, similarly tagged. This demonstrates that different strains, even when genetically very similar, have different competencies, in terms of survival, when introduced into the environment as inocula. Interestingly, ribotype A not only proved to have exceptional survival ability, but when co-inoculated, improved the persistence of the poorly surviving transient isolate. In an attempt to determine which traits distinguished competent strains from transient forms, the authors examined a range of phenotypes that had previously been associated with rhizosphere competence, including the presence of flagella and lipopolysaccharides (de Weger et al., 1987a,b). However, the only trait found to distinguish the two populations was their organic acids utilization characteristics. Contrary to expectation, ribotype A utilized a narrower range of organic acids than the transient form, but grew faster on compounds they could assimilate (e.g. citric and lactic acid). The importance of nutrients in determining inocula performance, in the rhizosphere at least, is not surprising, in particularly organic acids, as they are significant constituents of the root exudates (Rovira, 1969). However, it is surprising that the range of organic acids utilized by ribotype A was narrower than the less competitive strains. Indeed, the ability to utilize a diverse range of nutrients has previously been correlated to competitive ability in the rhizosphere (Bakker et al., 1993; Oresnik et al., 1998). The authors concluded that, in some cases at least, the ability to grow rapidly on a more narrow range of substrates can confer a competitive advantage, if that compound is abundant in the habitat. However, the contrasting conclusions of these studies highlights the difficulty of pin-pointing the specific traits that confer fitness on inocula, which are very much dependent on the characteristics of the habitat and strains, and their interactions, which in turn are unlikely to remain constant.

In a similar study, the authors also noted that pseudomonad strains had different rhizosphere competencies, and similarly used a dual selection criterion. In this instance numerical dominance in the target habitat (rhizosphere) and the ability to degrade a contaminant (naphthalene) were used to select strains for long-term phytoremediation applications (Kuiper et al., 2001). The authors demonstrated a 100-fold increase in strain survival after two enrichment cycles within the Lolium multiflorum rhizosphere, in addition to naphthalene degradation. The exceptional root-colonization ability and excellent performance of the selected strain was attributed to its ability to effectively utilize the nutrients specific to the plant cultivar selected, and the careful and systematic targeting of the strain.

In a series of studies, a systematic selection/isolation programme was undertaken whereby bacterial inocula were developed to treat spent metal working fluids (MWF) in bioreactors (van der Gast et al., 2003a; 2004), also based on significant criteria that reflected key features of the microbial habitat. However, in this instance the authors used a triple-selection criterion. These were: (i) the relative abundance of source populations in the target habitat (waste MWF), (ii) tolerance to co-contaminants (MWF are chemically mixed) and (iii) the ability to degrade components of MWF. Extensive analysis of the microbial community composition and structure of a single MWF formulation, both temporally and spatially at a worldwide scale, were undertaken to identify ubiquitous populations in waste MWF (van der Gast et al., 2002; 2003b). Subsequent screening cycles were based on the ability of isolates to tolerate the toxicity of co-contaminants and to catabolize individual chemical constituents. A consortium of four isolates was generated that proved to be 85% more effective at processing waste MWF in bioreactors than undefined inocula from sewage. It also demonstrated exceptional persistence in the chemically dynamic bioreactors. Perhaps even more importantly, in terms of applications, the consortium performance was more consistent and predictable than those from uncharacterized microbial communities (activated sludge), a feature of great value to chemical engineers and technicians who design and manage such systems.

Improved exploitation of robust strains

Over a decade has passed since it was first proposed that inoculation of strains containing mobile genetic elements (MGE) might provide a mechanism for the introduction of specific traits into microbial communities (Brokamp and Schmidt, 1991; Fulthorpe and Wyndham, 1991). For instance, the transfer of 2,4-dichlorophenoxyacetic acid (2,4-D) degradation plasmids pEMT1 from an introduced donor strain to the indigenous soil microbial community resulted in the complete degradation of 2,4-D after 19 days, while the contaminant persisted in the non-inoculated soil after 89 days (Dejonghe et al., 2000). Enhanced 2,4-D degradation was attributed to the emergence of transconjugants (105 colony forming units per gram soil), as the donor was undetectable before degradation started. It might be argued that this approach would negate the requirement to take the trouble to screen for competitive strains, as we are advocating, as the donor is no longer required, once catabolic activities are transferred and expressed in the indigenous bacteria. However, it can be equally reasoned that introduction of robust strains, containing desired genes, that persist long-term in a habitat, would have greater opportunities to transfer the catabolic genes. Indeed, plasmids-borne catabolic genes (e.g. 3-chlorobenoate and 4-chlorobiphnyl) have been successfully introduced, by conjugation, into strains specifically selected for their tolerance (Top et al., 2002). For instance, introduction of catabolic genes into desiccation-tolerant strains resulted in good degradation of the contaminants (biphenyl, isopropylbenezene and 3-chlorobenzoate) by the tranconjugants in desiccated soil conditions, and the strains demonstrated excellent shelf life when stored in a dried form (Weekers et al., 1999).

Enhanced degradation rates of carefully selected adapted strains have also been achieved by recombinant approaches. Watanabe and colleagues (2002) screened a bacterial community in an activated sludge plant for phenol degradation, and selected the dominant (most abundant) population Comamonas sp. rN7. Phc genes encoding for phenol hyroxylase, from the parent strain C. testosteroni R5, were integrated into the chromosome of the Comamonas sp. rN7. The specific phenol-oxidizing activity of the resultant transformant, designated rN7(R503), was three times greater than the activity of the original strain R5. Quantitative PCR revealed that the phc genes were retained in the transformant, at 108 copies per ml of liquor, for more than 35 days, while the parent R5, the original source of the phc genes, was undetectable. Furthermore, inoculation of phenol containing activated sludge with the transformant resulted in high phenol-oxygenating activity and improved resistance to phenol shock loading, compared with inoculation with the parent.

Taking a similar approach the catabolic genes were genetically engineered into persistent and adapted Deinococcus radiodurans strains, specifically selected for their natural tolerance to high doses of radioactivity, and modified to express the cloned mercury (II) resistance genes (merA), derived from Escherichia coli (strain BL308) (Lange et al., 1998). The resultant was a strain able to grow in the presence of radiation, a natural characteristic of this organism, and ionic mercury, which it transformed to the less toxic elemental mercury (II). Similarly, the same radiation-tolerant strain was modified to express toluene dioxygenase (tdo) activity. Cloning of the tdo genes into the chromosome of D. radiodurans imparted the ability to oxidize toluene, chlorobenzene and indole. The recombinant was also capable of growth and functional synthesis of tdo in highly irradiated conditions (Lange et al., 1998). This approach puts us in very strong position not only to select persistent and active strains, but to add functionality to well performing strains.

An alternative but complementary approach, successfully used to maintain the catabolic activity of introduced inocula, is to manipulate key environmental parameters selected specifically to favour catabolic expression and survival of introduced cells. Again this approach requires good understanding of the ecology of the selected strains. For instance, Pseudomonas stutzeri KC was selected for site inoculation because of its ability to transform carbon tetrachloride to carbon dioxide (Dybas et al., 1998). Previous studies revealed that the inoculum was favoured by anoxic conditions and demonstrated an exceptionally high affinity for iron (Criddle et al., 1990). By manipulating the pH in the field to pH 8, which lowered iron availability, the strain was able to persist and actively degrade the contaminant, by out-competing indigenous populations that were unable to obtain adequate concentrations of iron.

Future perspectives

It would be naive to believe that by simply picking the ‘right’ organisms or manipulating the right field parameter, bioaugmentation will suddenly become as reliable and predictable as engineered systems. Inevitably, as with most biological systems, there will always be an element of unpredictability. Such biotic factors as inoculum density, preparation and modes of introduction are known to greatly influence performance (Vogel and Walter, 2001). It is not practical to tailor consortia specifically to each habitat and eventuality. However, it is clear that members of the same genera/species do not all have equal fitness and some are likely to be competitive in a broader range of scenarios, while others may be more suited to specific conditions and habitats. With improving knowledge we should become more effective at determining and targeting the kind of organisms or traits (e.g. resistance to low pH or high heavy metal concentrations, poor nutrient conditions or high salinity) that are likely to suit specific conditions and remedial requirements. This should then allow us to develop a suite or culture collection of well-characterized organisms which maintain exceptional catabolic ability and tolerance in a broad range of chemical and environmental stresses. These superbugs are referred to as an Heirloom strains – named after the microorganism's often long lineage within research groups, where they acquire an ever-increasing dossier of desirable traits from an equally long list of researchers (Singer et al., 2005a). For instance, bioaugmentation of soils co-contaminated with organics and metals may see greater success with the application of microorganisms selected specifically for their dual catabolic competence and tolerance of toxic metals. Furthermore, such a collection would avoid the impractical requirement of painstakingly selecting and tailoring inocula for each specific site.

The notion that different types of microorganisms have different survival strategies and opportunities is far from new, but seems to have been largely forgotten. The great pioneering microbiologist Winogradsky distinguished between two classes of microbes in terms of their survival strategy (Winogradsky, 1924). He coined the term ‘zymogenous’ microbes (e.g. Pseudomonas), for those microorganism that exist mostly in a resting phase, which demonstrate brief periods of activity, stimulated by the appearance of an available substrate, returning to quiescence, when resources were exhausted. In contrast, the ‘autochthonous’ forms were defined as those organisms that demonstrate slow but continuous activity (e.g. Arthrobacter). Later, Hirsch and colleagues (1979) utilized the term ‘r-strategist’ to refer to an organism typified by a high Km and Vmax, suggesting the need for a large food source to induce growth followed by a high metabolic rate and subsequent short-lived but large increase in population size. ‘K-strategists’ are typified by a low Km and Vmax, and an equivalent metabolic and population growth response to both large and small nutrient introduction (Hirsch et al., 1979). On this basis, if a strain for bioaugmentation was required that could survive long-term in a habitat, slowly and continuously degrading a contaminant, an autochthonous candidate would be most suitable. Zymogenous forms would be ideal when there were occasional pollution episodes. When the contaminant was present, the zymogenous strain would explode in activity and growth rate, their numbers then rapidly diminishing when the contaminant was completely degraded. These scenarios may seem far fetched, and they probably are, but they serve to stress the point that it is essential to select the right ‘strategist’ for a specific task. However, what is described above may not be too far from the real world. In a long-term study into the persistence of bacterial inocula, the performances of Arthrobacter (A109) and Flavobacterium strains (P25), selected specifically to represent the two contrasting classes of microorganisms defined by Winogradsky, were evaluated in the field. As Winogradsky might have predicted, after 87 days, counts of the zymogenous form (P25) fell below detectable limits, while those of the autochthonous form (A109) were detected in excess of 300 days after introduction, when monitoring ceased.

Even though Winogradsky's theory of survival may now seem simplistic and may not survive the test of time, or the scrutiny of modern techniques, it does at least provide some theoretical basis on which we can more systematically select strains for exploitation for specific tasks in bioaugmentation. With the introduction of new technologies, we are in a much better position to understand the bewildering extent of microbial diversity, and to develop ecological theories and models (Curtis et al., 2003) that will inevitably help target the best microorganisms for any one task. Moreover, by minimizing the size of the ‘black box’, we are better able to respond to changes in system and environmental parameters – ultimately leading towards a more robust and reliable bioaugmentation.