Roundtable: 15 Years in Pharmaceutical Microbiology

Donald Singer
Global Lead Quality Manager, Microbiology
GlaxoSmithKline

Peter Noverini
Field Applications Scientist
Azbil BioVigilant, Inc.

Mark W. Kaiser
Director, Microbiology
Eurofi ns Lancaster Laboratories

Luis Jimenez, Ph.D.
Biology and Horticulture Department
Bergen Community College

Tim Russell
Field Market Developer
TSI Inc.

Priya Balachandran, Ph.D.
Product Manager
Life Technologies

Renaud Jonquieres, Ph.D.
Vice President, Pharmaceuticals and Cosmetics Franchise
bioMérieux

How has the pharmaceutical microbiology arena progressed over the last 15 years? note some major hurdles and milestones.

DS: Fifteen years is actually a short time span for changes in microbiology. Yet, we’ve seen an increase in considerations for alternative methodologies (rapid micro). Suppliers of new technology have begun to ask customers how to apply their technology and have revised their instrumentation packages to fi t many diff erent environments.

In Microbiology, some milestones I can note:

• Regulatory scrutiny of sterile products manufacturing has continued to increase and has led to guidance requirements for risk rationale; for example, the rewrite of 21CFR612 for sterility testing, the FDA Guideline for Container Closure Integrity in lieu of Sterility Testing, the MHRA ‘GMP’ Annex 1 pre-fi ltration bioburden action ‘limit’, and the USP <1116> contamination recovery rate recommendation
• Alignment of regulatory guidance with ICH Quality standards Q7, Q8 and Q9 which impact microbiological risk
• Increased published articles and books about pharmaceutical microbiological control
• Forums and media for communicating microbiological issues and solutions have increased Some hurdles I have seen are:
• Diffi culty in the implementation of new alternative methodologies and concern regarding ‘equivalency’ to pharmaceutical standard methods
• Economics leading to Microbiology labs to lean staffi ng, less capabilities, less time to innovate, and concern for improvement of supervision of analytical teams that lack experience or suffi cient knowledge in the fi eld
• Increase in instrument-based technology jumped ahead of adequate education for users to make good scientifi c decisions
• Continuation in pockets of the industry where management lacking scientifi c knowledge oversee microbiology operations

PN: Pharmaceutical microbiology has seen valuable incremental advancements, but has not matched the evolutionary pace of other areas such as manufacturing technology or statistical- and qualitydriven design. The creation of microscale, disposable-based manufacturing lines for clinical drug trials and widespread Six-Sigma belt training are examples of peripheral evolutions while strides in microbiology have been mostly limited to efficiency improvements. While it could be argued that major manufacturing advancements like isolator and RABS systems are, in part, driven by the microbiologically-derived imperative to separate people and product, the true potential of these new systems is often hampered by old microbiological paradigms (e.g. needing to insert/collect/extract a growth-based settling plate within an aseptic isolator).

To be sure, advancements such as the Analytical Profile Index method of rapid microbial characterization and quantified, lyophilized pure-strain cultures are among notable microbiological milestones. However, the backbone of microbiological analysis continues to focus on methods tied to crude quantification tools and metabolic analysis using stains such as the Gram stain. These methods pioneered by Pasteur, Bergey, and others a century ago play an important part in pharmaceutical microbiology, but the tendency to limit one’s microbiological worldview to what has come before is a major hurdle to the adoption of better microbiological methods.

MK: A major milestone in the area of sterile product microbiology is the use of isolator technology for sterility testing. Fifteen years ago, isolators were relatively new, and with the exception of larger pharmaceutical firms, most sterility testing was performed in cleanrooms. There was a strong focus on positive rates, and the financial impact of these was significant. While the use of isolators does not guarantee the elimination of false-positive results, in laboratories that understand the practical limitations of the technology (glove integrity for example) and develop procedures to overcome these limitations, positive rates as low as 0.02% and approaching 0 have been demonstrated.

PB: A lot of changes have been made over past 15 years. In pharmaceutical microbiology, where it’s a more regulated environment, results are used to make important decisions. Traditional microbiology (culture-based) techniques like morphology, gram staining, and biochemical tests are quite prevalent as a first screening step, followed by other methods to confirm microbes. One of the milestones was the shift to modern molecular methods as the confirmatory method for identification. Modern molecular methods provide quick turnaround, more accuracy and faster results. They enable establishment of robust environmental monitoring programs and rapid tracking of contaminating microorganisms.

The other milestone was the development of software to support rapid data analyses. In the past, traditional microbiology determined results by looking at a plate of bacteria. You would look at morphology and phenotype under a microscope. Then, you allow the bacteria to grow, and record information for identification. Current methods, including phenotypic methods like MALDI and sequence-based MicroSEQ® ID systems, provide sophisticated analyses protocols with reference databases that allow a great deal of information about microorganisms being identified (by either genotype or phenotype). We’ve moved from a notebook recording to an automated workflow, which includes instruments that automate results, match with existing data, and give you quick and accurate output.

RJ: While science and technology have progressed at an unprecedented pace in the academic area, practices remained relatively stable within the pharmaceutical industry. Traditional microbiology method still represents the vast majority of tests performed. Microbial techniques have improved with automation, superior packaging, improved formulations, greater precision and enhanced ease of use for the lab but the basic methodologies of microbiology– many of which are manual – remained the same. The early push for more rapid methods began in Analytical Method Development labs. Few projects went live or were implemented in manufacturing. Still, a few pharmaceutical companies pioneered the use of rapid methods in the early 2000s. Genzyme received first FDA approval to release one of its key biotech drugs with a rapid microbiology method in 2004. Since then, several companies communicated on implementation of Rapid Methods, including for example GSK, Novartis, Schering Plough and Alcon. The 2008 financial crisis slowed investment in new technologies and equipment. Nevertheless this time of scarcity helped to further reveal areas for improvement in Microbiology leading now to a dynamic trend of “Efficient Microbiology” where Microbiology becomes a tool for Lean operations and science-based decisions rather than being considered as a cost center.

Pharma users are slowly but surely turning to Rapid Microbiological Methods, what are the drivers for this and why the delay in adoption?

DS: I mentioned a ‘hurdle’ earlier of implementation difficulties. As scientists, we are anxious to use new technologies. Drivers are the excitement of implementation and performing our 100 year old methodology in a new and different manner. Yet, our biology conscience tells us to be aware and be careful, and develop a better understanding of the differences in results that may occur with the new methods. We have grown to understand that variation in biology is the norm, and we expect results to show mostly probability in occurrence based on sampling accuracy or representativeness. We also have learned that there are better controls for microbiological contaminants in the formulation, compounding, manufacturing and packaging of every drug product than trusting laboratory tests. So, if we can accept the latter holistic approach to microbiological control, then we can release our tight scrutiny and expectations of the new methods from being ‘equivalent’ to the highly variable classical methods, which can lead to allowing more innovative methods to be used routinely in our labs and manufacturing environments so we can develop the understanding and historical data needed to accept them as improvements to our testing tool box.

PN: The promise of RMM technologies includes many facets of operational excellence, depending largely on the particular RMM technology and application. Some RMM technologies focus primarily on the “R”, enabling increased manufacturing efficiencies simply by returning traditional, microbiologically-based results more rapidly. Other RMM technologies enable gains beyond incremental speed such as improved performance (sensitivity, precision, etc.), real-time situational awareness, lower cost, and higher quality (online, continuous, integrated data, etc.). These RMMs offer the capacity to go beyond simply testing to compendial requirements, driving operational excellence and value-added paradigms (e.g. real-time risk analysis, Parametric Release, etc.).

RMM benefits (and ROI) often appear proportional to the challenges in adoption. Rapid forms of traditional methods produce similar data streams (e.g. units of measure) that fit nicely within existing quality and regulatory paradigms. These same paradigms, however, can pose hurdles for RMMs that produce dissimilar data. Both regulatory authorities and industry experts have stressed the importance and ability to rethink old paradigms during RMM implementation. RMMs have an important role in pharmaceutical manufacturing, and difficulty in replicating or augmenting a compendial test should not dissuade implementation. Having internal end-user champions and a willingness to challenge old perspectives are keys to successfully realizing RMM potential.

MK: A number of factors impact adoption of these technologies. There have been minimal changes to the microbial compendial methods in the past 20 years. The pharmaceutical industry is conservative, and the potential of delaying product approval due to the use of a new technology poses a financial risk to the company. Cost of the rapid system combined with potential product obsolescence is also a concern.

More recently, regulatory agencies have demonstrated a greater willingness to accept rapid technologies and encourage the adoption of these technologies when appropriate. One example is encouragement of the vaccine industry to implement rapid sterility methods to support testing vaccines for pandemic flu strains.

Growth-based, nondestructive rapid methods with the appropriate sensitivity are available today while greater clarity around the expectations for demonstrating equivalency between the rapid method and compendial gold standard method make implementation possible. For example, a rapid sterility equivalency study that utilizes a growth-based, nondestructive method that produces results in seven or less days, detects stressed organisms in inhibitory matrices and has a demonstrated detection limit of 1 CFU, has a strong possibility of achieving regulatory acceptance.

TR: One driver of RMM is regulators wanting to see pharma adopt 21st century best practices and technology. The US FDA PAT Initiative encourages innovation and ICH Q10 PQS mentions innovation numerous times. In this sense, RMMs are a perfect fit. Pharma users are actively examining RMMs as a way to be competitive, due to their potential to immediately provide improved efficiencies, enabling time and cost savings while maintaining patient safety.

Delays in adoption are due to many factors, such as regulatory authorities encouraging RMMs but not detailing them in aseptic processing guidance. Additionally, smaller companies do not want to be the first to adopt and are waiting for the large pharma companies to lead the way. Conversely, some larger companies do not want everyone to know when they adopt RMMs preferring to maintain a competitive advantage.

Delay is also inevitable when a new technology has the potential to be more sensitive than existing methods. Existing active air sampling and culturing could not be validated if it was new technology being introduced today. Similarly, there is not a single new method which is perfect. Therefore, it is important to understand the strengths and limitations of any measurement technology, new or old. Laser Induced Fluorescence, used to detect airborne viable particles in real-time, is an example of a new RMM. It is a different from the culture-based method and as such is going to give a different result.

PB: The key drivers for turning to Rapid Microbiological Methods (RMMs) are needs for accuracy and short time-to-result. Genotypic RMMs are famous for their ability to deliver this. The DNA sequence of the organism is not influenced by external factors. Genotypic methods allow more confidence and greater accuracy. The customer gets answers in a shorter period of time. In general, in big facilities with high throughput, these methods are attractive, because they allow for workflow automation.

Traditional methods often continue to be used because they are familiar, with inexpensive running costs, and provide good enough results in many scenarios. In addition, RMMs generally require a change in the way microbiology is routinely conducted and a certain amount of knowledge in molecular biology methods. When a greater need for improved accuracy of results is realized, that’s when a lab moves from culture-based methods to modern molecular methods.

RJ: Nobody wants to wait fourteen days to get an answer. Would we accept such a delay when searching the Internet? Pharmaceutical microbiologist are like everybody, they want answers immediately so that they can make the right decisions. These decisions help improve patient safety and drive, in a scientific manner, the efficiency of manufacturing processes. So while it seems like adopting rapid micro methods should be a no-brainer, this move is happening more slowly than expected. First of all, microorganisms are extremely diverse and many are still completely unknown. To be widely adopted, a method should provide a one-size-fits-all protocol detecting the largest variety of bugs. Very few techniques provide such benefits. Moreover the cost associated with alternative methods is often handled by the Pharmaceutical microbiologist whereas the benefits and return on investment are actually beneficial to the manufacturing process. By building transversal teams in the pharmaceutical industry, the benefits are better understood and the costs are easier to justify.

Discuss reasons for microbiological-related recalls in recent years as well as key prevention efforts.

DS: A colleague, Scott Sutton, co-authored (with Luis Jimeniz) a comprehensive review of recalls in an American Pharmaceutical Review article (Feb 2012). Generally, in the area of sterile products, recalls seemed to indicate that the products were recalled due to either a lack of adequate demonstration of validated sterilization practices or potential non-sterile conditions of packaging, such as pinholes or other defects. Adequate process validation and documentation are the best approach to preventing these types of recalls. The the concerns developed from compounding pharmacy recalls in the last two years are more complex than just process validation as a corrective measure. In the latter instances, better oversight and responsibility are important measures, along with separation and understanding the differences between GMP manufacturing and good pharmacy compounding practices. One must realize that a pharma manufacturing operation has a Quality role for oversight of GMPs and product quality, whereas a compounding pharmacy may understand quality only as an attribute and not as a preventative activity. These troubling events took us by surprise, I think, and thus many stakeholders have begun to generate a more robust plan for improvement in this area, including the FDA, the USP and the APhA.

Packaging integrity is a sterile and non-sterile product attribute. Building a better understanding of how package integrity relates to product protection from microbial contaminants is an active field. Often, we have taken for granted the packaging components and processes used to seal packages. We’re learning more about weaknesses in the supply chain and challenges brought by more complex processes and package designs. If we continue to learn and be aware of potential integrity loss root causes, we will be more pro-active in our design for quality thinking. Be on the lookout for revised guidance from the USP (Chapter <1207>) and the PDA (revision of former TR27 document).

MK: The majority of recalls of non-sterile products are associated with objectionable organisms. While GMP regulations clearly define the requirement for absence of objectionable organisms in these types of products, there continues to be a lack of clarity on this issue. To avoid regulatory issues, companies need well-developed procedures to address objectionable organisms. The compendial microbial limit tests as written are not adequate to detect objectionable organisms, and additional modifications to the compendial methods are necessary. One approach is to streak the enrichments prepared in the compendial test to non-selective media in addition to the required selective media. The next step is to identify isolates and evaluate the organisms using a riskbased approach that considers a number of factors such as, dosage form, intended use, potential of the isolate to degrade the product and infection risk associated with the isolate.

LJ: Microbiological recalls have been increasing due to enforcement actions by federal agencies and recent contamination incidents where major violations of GMP practices resulted in morbidity and mortality cases [1]. Previous studies reported an increase in the number of product recalls by the FDA. The first study reported a disturbing trend in the numbers of products recalled by microbial contamination [2]. The study was published in 2007 and covered a period of time from 1998 to 2006. This was the first comprehensive study looking at the types of microorganisms found in product recalls reported by the FDA describing the types of products and microbial contaminants. Burkholderia cepacia was found to be the number one microbial contaminant in non-sterile and sterile pharmaceutical formulations. Yeast and mold were found to be responsible for 23% of the recalls in non-sterile and 7% of sterile products. Unfortunately, mold was found to be the reason for product contamination during a recent meningitis outbreak due to contaminated syringes [1]. Of the 193 recalls in sterile manufacturing, 78% were due to lack of sterility assurance.

A new study was published in 2012 with a different time range. Recalls were analyzed from 2004 to 2011 [3]. The findings demonstrated that the lack of sterility assurance was still a major problem for sterile products. However, the data showed a significant increase in the numbers of recalls from 2009 to 2011 [Figure 1, reference 3]. B. cepacia remains the number one microbial contaminant in recalls for non-sterile and sterile products. Although B. cepacia is an opportunistic pathogen for immunocompromised patients, the metabolic capacities of these bacterial species have been severely underestimated in pharmaceutical quality control [4]. B. cepacia is capable of growing on nitroaromatic and aromatic compounds by the action of different enzymes such as monooxygenases and dioxygenases [5]. Therefore the health hazard to patients not only makes B. cepacia a real nightmare for quality control microbiologists but these bacterial species can compromise product stability and purity by degrading active ingredients and excipients resulting in sub potent formulations. Many pharmaceutical formulations are based upon nitro aromatic compounds [5]. For instance, antipsychotic and analgesic drugs are based upon aromatic structures sensitive to biodegradation attack by mono and dioxygenases from microbial contaminants. B. cepacia posses a diverse genotypic and phenotypic potential to break down pharmaceutical active ingredients and excipients [6].

Further discussions by industry and regulatory agencies will provide better conditions to control microbial contamination and reduce the incidents of recalls, morbidity, and mortality by optimizing training, regulatory guidance, and enforcement.

RJ: In addition to the unfortunate and rare cases of improper manufacturing practices, increased scrutiny and precaution – driven both by regulatory concerns and the manufacturers themselves – certainly account for a large portion of recalls. Among these precaution principles, the fear of Objectionable Organisms is key. The highest number of recalls incriminates B. cepacia, an organism that is resistant to many traditional antimicrobials and responsible for severe infections in immuno-compromised patients. This organism is absent from the list of compendial specific bugs to detect, thus it is not tracked systematically. The use of modified Total Viable Count testing designed to specifically detect such an organism could help to improve the safety of drugs and limit the number of recalls.

Describe challenges of maintaining GMP and the advantages these guidelines present.

DS: Briefly, and with passion, I can say that microbiologists view GMPs with a more focused perspective. Microbiological control must be robust, consistent and appropriate for the manufacturing operation. We are learning more about the relevance of microbiological monitoring to product quality. Both the activity of monitoring and the intent of a controlled environment are related to GMPs. My view is that we should be designing and implementing a microbiological control program based on patient and product risk. This should be sufficient for a GMP-compliant environment. Yet, there are still a few influences in the regulatory field who continue to portray monitoring programs as a means of building GMP compliance in the name of ‘product quality’. We should be cognizant of this misdirected influence, and continue to improve and convey that a well-understood manufacturing process and environment is only partially based on monitoring data and mostly based on good design. In fact, maybe we should coin another industry acronym , ‘GMDP…good manufacturing design practices’; this could be more advantageous to improvement strategies.

What are the complications associated with control of microorganisms in biopharmaceutical manufacturing? How are these addressed?

DS: In simple terms, one could think that the use of growth media, live cells, and large equipment with optimal microbial growth conditions can either lead to optimal production of an intended biological substance or a disastrous contamination event.

Our knowledge is catamount to ensuring that we prevent the latter type of event. In fact, biopharmaceutical manufacturing has brought to us a continuum of learning with regards to microbiological control. Since many of these biological substances cannot tolerate classical ‘destructive’ terminal sterilization methods, we have seen development of the most robust designs for prevention or removal of microorganisms in an upstream process. The techniques used to remove microbial cells, bacterial endotoxin and viral contaminants are commonly understood by biopharm manufacturing scientists and provide the customized design parameters for microbiological control. Challenges of biological substance stability have led to more awareness of package integrity as it relates to impact during cold storage conditions. The hold time of a bulk substance preparation during manufacturing is scrutinized closely as it relates to the growth potential of contaminants. Viral contaminants and mycoplasma are also ongoing concerns for removal or inactivation. This wide spectrum of microbiological contaminants make it mandatory to design a biopharmaceutical process with holistic thinking for control, to assure purity and safety of products.

PB: Bioburden is omnipresent no matter how sterile the environment is. Quick and correct identification of problem microorganisms is therefore crucial to contain contamination events. Having proper systems in place, such as environmental monitoring programs can have a big impact on potentially problematic situations such as loss of large volumes of samples facility shut down in event of contamination. An accurate and rapid microbial identification method that is set up within the facility also supports efficient control of microorganisms.

RJ: Biopharmaceutical manufacturing requires living organisms for the manufacture of biological drugs such as monoclonal antibodies or Enzymes. In order to produce such high-value biologics the conditions must be highly favorable to living organisms, meaning nutrient sources such as oxygen, energy and, of course, water must be abundant. As such, unwanted microorganisms love to take advantage of this friendly environment. The length of these aseptic processes reaching up to two months and the absence of final sterilization method make biopharma manufacturing one of the highest risk processes for microbial contamination.

In order to prevent bad surprises at the end of the process, regular in-process controls are needed. Mitigating the risk by early and systematic analysis provides a rationale to move safely from one step of the process to the next. Alternative methods are tools of choice to provide such valuable information. They do not aim at replacing final compendial testing but to complement it. Microbial controls should target product and raw material bioburden, but also environment. Robust sampling plans allow for the definition of a clear baseline and help the team take action before systemic failure occurs.

What are your expectations for the next 15 years in this field?

DS: I believe we are just beginning to accept the potential of alternative technologies for routine /monitoring use as in-process tools, rather than release methods. If suppliers can make these tools more economical and if industry users can lower their expectations for level of ‘validation’ scrutiny, there will be a flurry of implementation over the next 10 years. Since we are already in an era of digital technology, I can see us routinely using tablets and bar-coding in 10-15 years to track environmental samples (plates), record results and trend the data in a mobile format. Some companies have begun this journey. Another interesting area that may need at least the next 15 years for acceptance is development and understanding of nonclassical means of terminal sterilization for materials that are sensitive to the current strict classical approaches.

PN: I believe a major theme in biopharmaceutical manufacturing will be increased automation, following other tech sectors with highly controlled processes/environments. At least two diverging philosophies describe the role of microbiology within this evolution. Some believe that, as automation expands, so as to virtually eliminate routine human interaction with the process/environment, an increasing reliance can be placed on parametric factors (laminar air rates, temp/RH, TOC, etc.) to ensure a system is under control and achieves a validated level of quality. This approach is aligned with tenets of Parametric and Real-Time Release initiatives. Another viewpoint is that microbiological measurements will always be necessary to demonstrate acceptable microbiological quality, and most sampling would necessarily require intrusion. Given established microbiological practices, these approaches seem mutually exclusive.

RMMs which offer data beyond simply faster compendial results, however, create an opportunity to bridge the philosophical gap. Some RMMs can be operated remotely, reducing the need for manual sample collection. Furthermore, online RMM systems can be integrated with process controls (e.g. HVAC system) to automatically maintain a predetermined state of control. An RMM-enabled future would reduce the need for operator interactions/interventions, allowing microbiologists to focus on interpreting the data and reacting with rapid microbiological tools when control cannot be maintained.

TR: In less than 15 years, real-time fluorescence based airborne viable particle detectors will be commonly used. Non-viable airborne particle counts as an indicator of viable contamination will no longer be necessary. The real-time detection of viable particles in liquids will be common place.

Today, the global regulatory expectation is very clear regarding classifying and monitoring of pharmaceutical cleanrooms for nonviable particles during aseptic processing. Current guidance as to when to monitor for airborne microbiological particles is not so clear, there is no distinction between classification and monitoring. Culturebased methods will continue to compliment rapid methods although the guidance surrounding microbiological monitoring will be enhanced, such as an expectation to monitor continuously during critical aseptic processing.

The variability associated with microbial measurements will be openly accepted as new techniques are compared to the current compendial approach. A first step has been taken with the revision of USP 1116 recommending a transition from fixed low count limits to number of occurrences. This is an excellent platform for discussion regarding the significance of low microbial counts associated with the highly variable compendial method.

It is almost impossible to purchase a non-viable particle counter today and not have them correlate with each other. Conversely, it is almost impossible to purchase an active air sampler from different manufacturers and obtain comparable results. It is inevitable that this situation will improve through the update of the relevant ISO standards and their association to the GMPs.

PB: I see more shifts to molecular methods. Molecular methods, particularly sequence-based (including next-generation sequencing) methods will have all the advantages of the methods that are present today. Such technologies are well-positioned to address hurdles such as accuracy, fast turnaround time, strain typing and addressing mixed culture samples.

Regulatory agencies already identify genetic methods for identification of microbes as the preferred method. With the increasing adoption rate of RMMs, we are gradually seeing a shift in the way QC microbiology is being conducted.

RJ: More and more alternative methods will be made available from microbiology vendors in the coming years. They will continue to address today’s unmet needs and reduce non value-added tasks in the lab. If Pharmaceutical companies reinforce upstream collaborations with vendors it will allow key innovations to diffuse more rapidly such as Mass Spectrometry Identification or Automated Cytometry for example. Also, giving more flexibility to pharmaceutical companies to implement in process alternative methods would certainly help to leverage these technologies more quickly. Today’s expectations for validating any given alternative method are so high that I doubt many of the traditional and accepted methods used daily would pass such criteria. Simplifying the implementation for in-process tests while sticking to the compendial methods for release would allow for improved product quality and cost savings, all while keeping the reference framework of the traditional methods.

References

1. http://www.cdc.gov/hai/outbreaks/meningitis.html
2. Jimenez, L. 2007. Microbial diversity in pharmaceutical product recalls and environments. PDA Journal of Pharmaceutical Science and Technology 61:383-399.
3. Sutton, S.W., and L. Jimenez. 2012. A Review of Reported Recalls Involving Microbiological Control 2004-2011 with Emphasis on FDA Considerations of “Objectionable Organisms”. American Pharmaceutical Review 15, January/February: 42-57.
4. Torbeck, L., D. Racassi, D.E. Guilfoyle, R.L. Friedman, and D.Hussong. 2011. Burkholderia cepacia, the decision is overdue. PDA Journal of Pharmaceutical Sciences and Technology 65:535-543.
5. Kou-San Ju, and R.E. Parales. 2010. Nitro aromatic compounds from synthesis to biodegradation. Applied and Environmental Microbiology 74:250-274.
6. Lessie, T.G., W. Hendrickson, B.D. Manning, and R. Devereux. Genomic complexity and plasticity of Burkholderia cepacia. FEMS Microbiology Letters 144:117-128.