|Home | About | Journals | Submit | Contact Us | Français|
We examined the available evidence on the effectiveness of measures aimed at protecting humans and the environment against the risks of working with genetically modified microorganisms (GMOs) and with non-GMO pathogenic microorganisms. A few principles and methods underlie the current biosafety practice: risk assessment, biological containment, concentration and enclosure, exposure minimization, physical containment, and hazard minimization. Many of the current practices are based on experience and expert judgment. The effectiveness of biosafety measures may be evaluated at the level of single containment equipment items and procedures, at the level of the laboratory as a whole, or at the clinical-epidemiological level. Data on the containment effectiveness of equipment and laboratories are scarce and fragmented. Laboratory-acquired infections (LAIs) are therefore important for evaluating the effectiveness of biosafety. For the majority of LAIs there appears to be no direct cause, suggesting that failures of biosafety were not noticed or that containment may have been insufficient. The number of reported laboratory accidents associated with GMOs is substantially lower than that of those associated with non-GMOs. It is unknown to what extent specific measures contribute to the overall level of biosafety. We therefore recommend that the evidence base of biosafety practice be strengthened.
Work with pathogenic microorganisms and genetically modified microorganisms (GMOs) requires precautions that guarantee the safety of humans and the environment, including laboratory personnel, patients treated with GMOs, and other persons who could be exposed to these microorganisms. During the past decades, responsible authorities and researchers have therefore developed regulations and guidelines that in some detail describe containment measures and working instructions (Fig. (Fig.1).1). For GMOs such regulations and guidelines appear to be largely derived from those developed for working with the natural, genetically unmodified pathogenic microorganisms from which these GMOs have been derived.
Despite containment measures and guidelines, laboratory infections, usually involving non-GMOs, occur more or less commonly, suggesting that biosafety rules are not always effective or complied with. The guidelines and instructions for working with GMOs appear to be largely effective, as there have been no major accidents with GMOs or with their unintended release. Nonetheless, despite such regulations and the lack of major accidents with GMOs, there appears to be continuing concern about the health and safety of individuals and the environment exposed to potentially hazardous GMOs (65). It has also been noted that the laws and regulations governing the biotechnology world are outdated, are not comprehensive, and span too many agencies (65). Indeed, while the natures of the risks and the measures to handle these risks are largely identical for GMOs and non-GMOs, in many countries there are different regulations for GMOs and non-GMOs. This may be because the latter were presumed to carry greater risks for causing ecological disturbances upon unintended release. For example, in The Netherlands, the Ministry of the Environment oversees work with GMO, while the Ministry of Social Affairs oversees work with human pathogens. The different regulations and overseeing authorities may be confusing to workers in the field. Moreover, it is unknown to what extent specific factors contribute to a safe biosafety practice. Thus, it is often unclear if and to what extent measures aimed at providing biosafety are based on documented evidence of their effectiveness.
A central question in this study is whether containment measures are effective and evidence based. One may argue that the evidence for the effectiveness of containment measures is at best indirect, i.e., based on the lack of many overt laboratory-acquired infections (LAIs). In addition, we could question whether the criteria to judge effectiveness are sufficiently developed. Indeed, the objectives of containment measures are often not explicitly defined, and without (quantifiable) objectives, evaluation of effectiveness is difficult. Furthermore, in finding evidence for the effectiveness of biosafety measures, it is important to judge the quality of the evidence. For comparison, in evidence-based medicine, systematic reviews, hypothesis-driven controlled laboratory experiments, and prospective studies provide a higher quality of evidence than case reports and expert opinion.
In this review we give a brief historical overview of the development of the current biosafety practice, we will try to identify which principles and methods appear to underlie it, and we will describe this current biosafety practice. We then present an approach for evaluating the effectiveness of biosafety measures to contain pathogenic microorganisms, and finally we summarize experimental and observational data on the effectiveness of containment measures. Our primary goal is to evaluate the evidence-based containment measures for GMOs. However, because such measures are largely based on containment measures for non-GMOs and because more data, although scarce, are available for non-GMOs, we also examine evidence-based measures to contain non-GMO pathogens. These data may be extrapolated to GMOs. We thereby hope to contribute to a conceptual framework that helps in further developing an evidence-based biosafety practice.
While Robert Koch had already developed some kind of biosafety cabinet (BSC), A. G. Wedum of the U.S. Biological Research Laboratories at Fort Detrick, MD, can be regarded as one of the pioneers in developing biosafety measures after the Second World War. He evaluated the risks of handling hazardous biological agents and developed practices, equipment, and facility safeguards for their control (113, 147, 148). Following his initial work, it is now regarded as conventional wisdom that enclosure and ventilation of contaminated work areas are important factors in eliminating LAIs. Besides safe microbiological techniques, primary barriers (safety equipment and personal protective equipment) and secondary barriers (facility safeguards) are now regarded as vital elements of containment measures.
It was recognized early that examining LAIs could be informative about the risks involved with laboratory work. Several comprehensive reviews of LAIs have therefore been compiled (116, 131). These early examinations recognized that the primary route of transmission of many of the causative agents was by aerosol, and they led to the development of laminar-flow BSCs. Legislation and guidelines that were introduced over the years have probably reduced but not eliminated the risk of occupational exposure to infectious agents.
With the growing ability to manipulate DNA in the mid 1970s, there was also growing concern about the potential hazards associated with recombinant DNA research and technology (9). GMOs could display the intended properties, but could also have unpredictable and undesirable features. There still appears to be debate about the health and safety of laboratory workers and animals, as well as the environment, exposed to potentially hazardous GMOs (65, 98). At the Asilomar Conference in 1975, general principles for dealing with potential biohazards related to GMOs were drafted. It was suggested that containment should be an essential consideration in the experimental design and that the effectiveness of the containment should match the estimated risk. Adjustment of the level of precaution to the level of risk would prevent infection without unduly impeding operations.
The first edition of the National Institutes of Health (NIH) guidelines for research involving DNA molecules appeared in 1976. Now, 3 decades later, there are a number of authoritative international guidelines, instructions, and recommendations for the safe handling and manipulation of hazardous biological agents, including GMOs. In 1984, the U.S. NIH and Centers for Disease Control and Prevention (CDC) produced the first edition of a guidebook, called Biosafety in Microbiological and Biomedical Laboratories, that is now considered a major reference text. The NIH/CDC and the WHO manuals are based on historical accounts of incidents with infectious microorganisms and extensive experience of experts working in this field, and they have been developed and improved over the last 30 years (152, 156). Legislation has been implemented, for example, in European Union and national regulations (145). In Europe, national authorities have based their regulations on directives from the European Union, such as the directive on the protection of workers from risks related to exposure to biological agents at work and the directive on the contained use of GMOs. A summary of important guidelines and manuals is given in Table Table11.
In judging the effectiveness of measures intended to ensure biosafety, it may be helpful to know which goals, principles, and methods of biosafety measures have been employed and to evaluate the scientific basis of their effectiveness. Because the natures of risks are largely identical for GMOs and non-GMOs, containment measures to handle these risks are largely identical for both. In the past biosafety measures have evolved step by step, and usually based on expert knowledge and experience, but without a unifying set of guiding principles. Explicit guiding principles are therefore usually lacking in most legal regulations and scientific papers. We here attempt to draft a hierarchy of such guiding principles and methods that may be applicable to both GMOs and non-GMO, as follows.
In the following sections we further elaborate on some of these principles and methods and on the way in which they have been incorporated in legislation and regulations.
From a thorough risk assessment procedure, considering all potentially harmful effects for humans and the environment, follows the risk classification of microorganisms and subsequent containment measures that should be taken to manage these risks. Both the nature and scale of activities need to be considered to estimate the possibility of exposure of humans and the environment and the consequences of such exposure. Examples of risk assessment procedures can be found at http://www.hse.gov.uk/biosafety/gmo/acgm/ecrisk.htm and in European Union Directive 90/219/EEC.
As indicated below, there are four risk categories for hazardous biological agents and four containment levels. Work with non-GMO microorganisms is usually done at the corresponding containment level. It is evident that a risk assessment for working with noncharacterized pathogens in a clinical laboratory may involve uncertainties. Through manipulation GMOs may acquire pathogenic properties that are unexpected and/or not well characterized or understood, necessitating a higher containment level than for work with the natural microorganism from which the GMO has been derived, or additional measures (see, for example, reference 145). Therefore, the nature of recombinant DNA sequences, vectors, and recipient organisms needs to be carefully evaluated, as well as any potential biohazard associated with particular experimental settings. It is particularly important to address whether or not genetic modification affects cell tropism, host range, virulence, or susceptibility to antibiotics or other effective treatments. Some considerations regarding the risk assessment and categorization of GMO activities are given in Table Table22.
A subsequent step in the selection of measures to ensure biosafety is to minimize biological hazards associated with the work by employing host microorganisms with a reduced host range, strains with natural or genetically modified characteristics that diminish their invading capacity or virulence, self-inactivating vectors, etc. Thus, employing biological containment is not restricted to work with GMOs. Natural pathogenic microorganisms may be replaced by less pathogenic microorganisms; for example, Mycobacterium tuberculosis may be replaced with nonvirulent mycobacteria, such as Mycobacterium smegmatis (27). Approaches for acquiring biological containment by genetic modification are given in Table Table33.
Examples of biological containment include the use of highly modified vaccinia virus Ankara, which has a significantly reduced pathogenicity, and the use of avian poxviruses, including canarypox and fowlpox, that have a restricted host range and do not replicate in mammals (93, 108). These viruses have a better safety profile than the classical vaccinia virus.
Biological containment for retroviral vectors has been obtained by providing gene products that are required for the production of progeny viruses, i.e., gag, pol, and env, in trans by packaging cell lines that stably express these trans-acting functions. Formation of replication-competent human immunodeficiency virus (HIV) is excluded when the env gene is missing from the packaging plasmid. Furthermore, the range of host species that can be transduced with such vectors can be manipulated by using particular env genes or by pseudotyping with envelope proteins of other viruses (29, 39, 118). Various envelope proteins are associated not only with a varying host range but also with varying stability and intrinsic toxicity. A concern has been that recombination events may give rise to replication-competent lentiviruses, but further modifications have been made to reduce this opportunity. These include the use of different transcriptional units, further attenuation by deleting nonessential genes, and self-inactivating vectors (17, 29). Self-inactivating vectors contain an inactive long terminal repeat, resulting in a lack of promoter activity and preventing potential transcriptional activation of (onco)genes downstream of the integration site. These vectors cannot be rescued by wild-type HIV. No replication-competent lentiviruses have been detected when such systems are used (29, 37). Altogether, the risk associated with the newly developed lentivectors is therefore minimal, while they have distinct advantages such as stable integration in nondividing and dividing cells, long-term expression of the transgene, and absence of an immune response. However, although viral infection using state-of-the art lentivectors is very unlikely, their high transduction efficiency may increase the risk of accidental exposure of lab workers, leading to a positive anti-p24 HIV antibody response and accidental transduction of potentially hazardous genes. Contact transmission thus should still be avoided.
Similarly, replication-deficient adenoviruses have been developed using a viral DNA vector and a helper cell line that has been stably transfected with the E1 region (E1A and E1B) of the adenoviral genome. Vectors prepared from the cell line lack the E1 region and remain replication defective (72).
Safe replication-deficient herpesvirus vectors have been developed by deleting the viral glycoprotein gD. Envelope glycoprotein gD is essential for virus entry but is not required for subsequent steps in the viral replication cycle. Phenotypically complemented gD null mutants can infect cells and can spread, both in vitro and in vivo, by direct cell-to-cell transmission. However, progeny virions released by the infected cells are noninfectious because they lack gD (109). Thus, an accidental infection remains restricted to a single round of replication. The same principle has been used for other viruses, for example, respiratory syncytial virus. Other ways to generate safe herpesvirus vectors are deletion of genes that are essential for virulence in vivo but that are nonessential in vitro or deletion of immediate-early genes that activate early and late gene expression and subsequent propagation of the crippled virus in complementing cell lines (68, 75). Replication-incompetent herpesvirus vectors have also been generated by the use of cosmid DNAs to provide the necessary viral gene products for propagation of defective viruses or amplicons. These approaches reduce the opportunities for generation of replication-competent viruses through recombination.
Safe vectors have been developed from the nonpathogenic adeno-associated virus. The vector lacks all viral genes and requires coinfection with a helper adenovirus or a helper-free packaging system (24, 120).
A high level of biological containment is also acquired by using the Autographa californica nuclear polyhedrosis virus, a member of the baculovirus family. These viruses normally replicate in insect cells but not in mammalian cells. Furthermore, by deleting the nonessential polyhedron gene, the virus becomes noninfectious for its natural host (106).
The classical example of a biologically contained bacterium is Escherichia coli K-12. E. coli K-12 is a debilitated strain that does not normally colonize the human intestine. The strain survives poorly in the environment and has a history of safe commercial use. E. coli K-12 is considered an enfeebled organism as a result of being maintained in the laboratory environment for over 70 years (154). E. coli K-12 is defective in at least three cell wall characteristics. First, the outer membrane has a defective lipopolysaccharide core which affects the attachment of the O-antigen polysaccharide side chains (26). Second, it does not have the type of glycocalyx required for attachment to the mucosal surface of the human colon (S. Edberg, unpublished report, U.S. Environmental Protection Agency, 1991) as a result of the altered O-antigen properties noted above. Finally, K-12 strains do not appear to express capsular (K) antigens, which are heat-labile polysaccharides important for colonization and virulence. K-12 thus is not able to recognize and adhere to the mucosal surface of colonic cells (26). The normal flora in residence in the colon thus can easily exclude K-12. Furthermore K-12 lacks other virulence factors (26, 46; Edberg, unpublished report).
Several approaches have been used to develop safe bacterial vectors for use as vaccines or for gene or protein delivery, in particular, strains of Salmonella, Shigella, Listeria, Mycobacterium, Vibrio, and lactic acid bacteria (71, 137). When developing vector vaccines, the challenge is to develop strains that are well tolerated by the recipient host and no longer persist in the environment, yet still induce protective immune responses. This is not always achieved, however (32, 141). Elucidation of biosynthetic pathways has led to the development of Salmonella vectors that were attenuated by disruption of genes encoding metabolic functions or genes located in a pathogenicity island (Salmonella pathogenicity island 2, which encodes a type III secretion system). Salmonella pathogenicity island 2 is required for survival and growth within macrophages (67). For example, attenuating mutations in Salmonella strains included msbB, galE, via, rpoS+, aroCD, htrA, cya, crp, cdt, asd, phoPQ, purB, sifA, and ssaV (33, 67, 84, 90). The best-characterized live attenuated salmonellae have mutations in the prechorismate pathway. These are the so-called aro mutants, which are defective in the production of chorismate, which is essential in the synthesis of aromatic compounds (90). It is evident that strains carrying different mutations differ in properties such as invasiveness and survival. To reduce the possibility of reversion to virulence, strains carrying at least two attenuating distantly located mutations have been produced (67, 141). Salmonella enterica serovar Typhimurium VNP20009 was developed to deliver potential therapeutic proteins to tumor sites. It was created by chromosomal deletion of two genes, purI (purine biosynthesis) and msbB (lipopolysaccharide biosynthesis) and was attenuated at least 10,000-fold in mice compared with the parental wild-type strain (84).
A promising Listeria vector vaccine is an L. monocytogenes auxotrophic mutant with deletions in alanine racemase (daI) and d-amino acid aminotransferase (dat) genes, two genes required for the biosynthesis of bacterial cell walls. The strain was highly attenuated in mice (142). The strain requires d-alanine to grow and survive. Another L. monocytogenes candidate vaccine strain (LH1169) contained deletions in actA and plcB, genes that are necessary for cell-to-cell spread and escape from secondary vacuoles, respectively (2). Deleting lecithinase activity in L. monocytogenes also results in inhibited cell-to-cell spreading (31).
The virulence of Vibrio cholerae is due mainly to the expression of cholera toxin (CT). Hence, strategies for attenuating V. cholerae for use as an expression vector for heterologous antigens have been to engineer mutants in which the CT gene (ctx) or the CT genetic element has been partially or completely deleted (71).
Stable mutant strains of Bacillus licheniformis, an industrially exploited species, were obtained by introducing defined deletions in recA and/or an essential sporulation gene (spoIV). These strains are totally asporogenous and severely affected in DNA repair, and they therefore are UV hypersensitive. In liquid media these strains grow equally well as the wild type. Hence, such genes appear to be suitable disruption targets for achieving biological containment (96).
Construction of a genetically modified Lactococcus lactis strain for intestinal delivery of human interleukin-10 (IL-10) employed a biological containment system by replacing the thymidylate synthase gene thyA with a synthetic human IL-10 gene. When deprived of thymidine or thymine, the viability of the strain dropped considerably, preventing its survival in the environment. Transgene escape through acquisition of an intact thyA gene is very unlikely and would recombine the transgene out of the genome. The system was validated in vivo in pigs (136) and was used in a gene therapy study (13).
Further improvement in enhancing the safety profile of bacterial vectors for gene transfer can be achieved by removing undesirable properties of plasmids, such as a prokaryote origin of replication and antibiotic resistance markers. These elements could lead to dissemination of prokaryotic replicative recombinant DNA. Darquet et al. (28) therefore developed so-called minicircles, which are supercoiled DNA molecules that lack such elements and contain only an expression cassette carrying the gene of interest. Furthermore, efficient suicide functions have been developed to ensure biological containment of bacteria (69, 128). Such systems achieve their goals when the GMOs self-destruct by expression of killing genes after fulfilling their jobs. Suicide systems are based on “lethal genes” that are triggered by preprogrammed conditions. Such systems, however, appear to differ in efficiency of the suicide function, and the less efficient ones may lead to selection of mutants that have lost their suicide function. However, no system can provide complete efficiency. One efficient system was based on the lethal E. coli relF gene, which prevents the transfer of plasmids to wild-type bacteria (69).
To enhance the safety of genetically modified yeast (Saccharomyces cerevisiae), genes encoding bacterial toxins have been used for containment control. Expression of the E. coli relE toxin gene was highly toxic to yeast cells, and this could be counteracted by expression of the relB gene (73).
In conclusion, we believe that the mechanisms underlying biological containment are usually well examined and understood. Evidence for their effectiveness is usually available, although only seldom in quantitative terms of infectivity and transmission. It is also important to note that not all GMOs are by definition biologically contained. They have reduced transmissibility or virulence only upon disruption of virulence factors. For example, very virulent recombinant influenza viruses and herpesviruses have been regenerated (30a, 47).
Central to biosafety programs is the concept of universal precautions (16). For that purpose, microorganisms are categorized into four risk categories on the basis of a risk analysis. Subsequent risk containment is focused not on specific infectious agents but on standard practices for handling infectious material that will prevent the transmission of all pathogens of that risk category. It is important to note that the principles, guidelines, and recommendations are basically the same for natural pathogens and GMOs. However, as would be logical, this conclusion did not lead to a single set of guidelines and recommendations for GMOs and non-GMOs, and thus there is some redundancy in guidelines and regulations.
The four categories for biological agents are based on their relative risk to laboratory workers and the community. In general, the following factors are considered in classifying biological agents: (i) the virulence of the biological agent or the severity of disease (in humans) (18), (ii) the mode of transmission (spread in the community and host range), (iii) the availability of effective preventive measures (e.g., vaccines), and (iv) the availability of effective treatment (e.g., antibiotics or antiviral drugs). The hazard of the infectious (non-GMO) agent increases from risk group 1, consisting of microorganisms not associated with disease, to risk group 4. Risk group 4 microorganisms can cause serious disease and can be readily transmitted, and effective treatments are usually not available (123, 152, 156). However, there are differences in the exact definitions as used by certain countries and/or organizations (such as NIH/CDC, WHO, and European Union) (Table (Table4),4), which result in differences in the exact listings of biological agents in each risk category (43; http://www.absa.org/XriskgroupsX/index.html). The main difference between the NIH/CDC classification and the WHO classification is that the latter includes hazards to animals and the environment. Another difference is that for risk group 3, the NIH/CDC states that “… therapeutic interventions may be available,” whereas the WHO and European Union state that “… effective treatment and preventive measures are available.” There also are differences in the description of transmission properties between the different classifications. In assigning an agent to a risk group, one must take into account that there are in all groups of microorganisms naturally occurring strains that vary in virulence and that may thus need a higher or lower level of containment (43). In general, regulators deal with this concept by taking the highest level of virulence into account.
There has been debate about the classification of particular microorganisms, in particular the Flaviviridae, variola virus, avian influenza A/H5N1 virus, and extremely drug-resistant M. tuberculosis strains, especially as to whether they should be categorized as category 3 or 4 biological agents (1, 20, 25, 40, 70, 97, 143). Because vaccination was stopped in the 1970s, variola viruses are now classified as category 4 biological agents. Although avian influenza A/H5N1 virus strains initially were classified as risk category 4 biological agents, susceptibility to antiviral drugs and the availability of effective vaccines may downgrade them to category 3 for further studies. Extremely drug-resistant M. tuberculosis strains (25) should be regarded as risk category 4 biological agent. Of note, to date no microorganisms other than viruses have been classified in category 4.
As for non-GMOs, GMOs are assigned to specific risk categories based on the risk assessement. For GMOs biosafety containment levels are assigned depending on the risk category of the donor organism, unless the modification may result in a higher or unknown risk. Therefore, the nature and function of insert sequences and the properties of acceptor microorganisms are considered. It is remarkable that definitions of properties of GMOs are sometimes quite strict (for example, the toxicities of vertebrate toxins, which are expressed as 50% lethal dose/body weight), while other properties of harmful gene products are not exactly defined. In the regulations, both the nature and level of virulence and transmissibility of GMOs are not well defined, at least not in quantitative terms (such as the basic reproduction ratio [R0] ). Sometimes host range is taken into account (for example, for baculovirus, ecotropic murine retroviruses, and papillomaviruses when they are used in nonpermissive host/vector systems). Properties may be unknown, such as the capacity of microbial DNA to integrate into the host genome (as, for example, for HIV) and the availability of vector organisms in the environment and therefore the possibility of persistence in the environment. A point for consideration is therefore the possibility of monitoring replication and survival of the GMO. In case of scientific uncertainty, the precaution principle is leading, resulting in higher categorization of the GMO or additional measures on a case-by-case base. Although the possibility of microbial transmission is taken into account, the possibility that transmission is reduced by herd immunity, either vaccine derived or not, is not mentioned explicitly by legislators. The assessment should include whether properties of inserted sequences can be expressed in the background of the host organism. Thus, gene-gene and gene-environment interactions should always be considered. As an example, IL-4 is not normally considered a harmful gene product, but when expressed by a murine ectromeliavirus, it drastically enhanced the virulence of this virus by inducing changes in cytotoxic T-cell function (58). Importantly, the regulations, being not completely detailed, largely function as a framework that must be elaborated by researchers and laboratory directors in their risk assessment, in the risk categorization of GMOs, and in the implementation of procedures and regulations in their laboratories (79). While the considerations in the risk assessment and subsequent risk categorization may show differences for GMOs and non-GMOs, many of the subsequent containment procedures are similar.
Phillips and Runkle (113) describe some principles underlying laboratory design. They introduce two concepts in designing laboratories. The first is the concept of primary and secondary barriers as described above, and the second provides the designers a logical division of major functional zones within a laboratory building. (Note that although elements of primary and secondary barriers are clearly recognizable, they are not mentioned as such in the European and Dutch regulations.) Phillips and Runkle (113) identify five functional zones in the facility (clean and transition, research area, animal holding, laboratory support, and engineering support). Primary containment measures minimize occupational exposure of laboratory workers. In addition to strict adherence to good microbiological practice, the primary containment barrier include physical separation of the biohazardous agent from the laboratory worker using closed vessels, personal protective equipment (e.g., gloves or full-body suits) and additional equipment (e.g., BSCs, enclosed centrifuge containers, or pipetting aids). The secondary barriers provide supplementary containment, serving mainly to protect other facility employees and to prevent the escape of infectious agents from the laboratory if and when a failure occurs in the primary barriers. These provide a separation between potentially contaminated areas in the building and the outside community. These measures may comprise special procedures (e.g., validated decontamination methods, training of personnel, strictly controlled access zones, interlocked doors, etc.) and special engineering and facility design features (e.g., decontamination equipment, showers, autoclaves, dedicated air handling system with filters, etc).
Secondary containment is very strict in high-containment laboratories. The high-containment laboratories (biosafety level 3 [BSL-3] and BSL-4) are airtight and have airlocks and a unidirectional airflow so that potentially contaminated air is kept inside. Thus, BSL-3 and -4 laboratories need to be negatively pressured, resulting in an airflow from adjacent areas into the laboratory and a filtered exhaust airflow outside the building without recirculation. Ten to 12 air changes per hour have been recommended, which removes approximately 99% of airborne particles in 23 min (92). Release of air into the environment is possible only through HEPA (high-efficiency particulate air) filters (rated 99.99% efficient with particles 0.3 μm in diameter and larger) or ULPA (ultra-low-penetration air) filters (rated 99.999% efficient with particles 0.12 μm in diameter). Filters are used for BSCs, autoclaves, incinerators, chemical decontamination showers, etc. In addition, HEPA filters are used to provide clean air to laboratory workers in full-body suits.
Biosafety containment levels have been categorized in a range from 1 to 4. As indicated above, the containment levels are assigned depending on the risk group of the microorganism (GMO or non-GMO) and the scale and nature of activities. Importantly, the BSL designations are based on a composite of the design features, construction, containment facilities, equipment, and operational procedures required for working with agents from the various risk groups (152). It is important to stress that although handling microorganisms of a certain risk group usually requires working at the accompanying BSL, a risk assessment should be made to take other specific factors into consideration. For example, particular experiments may generate high-concentration aerosols, requiring a higher degree of safety. Thus, professional evaluation, based on personal responsibility, should always guide the BSL for the specific work (152). Sewell (131) formulated some broad recommendations. BSL-1 is recommended for teaching activities with agents that are not associated with disease. BSL-2 practices are used in diagnostic laboratories that manipulate agents that are not transmitted via aerosols (e.g., hepatitis B virus [HBV], HIV, enteric pathogens, and staphylococci). BSL-3 is recommended when working with agents that are highly infectious and are transmitted via aerosols (e.g., M. tuberculosis, Brucella spp., and Coccidioides immitis) and for large-scale work with BSL-2 agents. BSL-4 practices are required when working with unusual agents that cause life-threatening infections for which no treatment is available.
Depending on country and/or regulation authority, there are differences between the exact requirements for each of the four containment levels, not to mention the sometimes confusing differences in nomenclature for the (high-)containment laboratories. Nulens and Voss (104) reviewed the basic practice, equipment, and facilities necessary for each of the BSL, based on European Union Directive 98/81 and the WHO biosafety manual. This information is summarized in a general way in Table Table5.5. A comprehensive listing of all requirements and equipment necessary for the four BSLs is presented in Table Table6.6. This list is adapted from the WHO biosafety manual (152).
At BSL-1, safety is achieved mainly by applying good microbiological techniques. To achieve a higher degree of containment and thus a higher degree of protection against LAIs, the number of requirements increases up to the maximum containment at BSL-4. Besides equipment, there are several codes of practice for laboratory access, personal protection, and working procedures. Scientific evidence for the efficiency of these measures is scarce, and most of these measures have been developed based on a long history of microbiological practice and common sense. Table Table77 describes the codes of practice regarding access and personal protection for BSL-1, -2, and -3 laboratories.
In The Netherlands, BSL-1 to -4 are called MLI to -IV for work with GMOs. These are based on the European Union Directives 90/219 and 98/81 and are implemented in national regulations (145). Many technical requirements and access and personal protection rules are similar, but a major difference is that according to Dutch regulations, at the MLII level (BSL-2) a class II BSC is not optional but is required. Another difference is that the Dutch regulations provide more detail with respect to procedures. For instance, it is stated, “prepare your work carefully limiting the necessarily movement from one place to another during the microbiological work. Collect all material and equipment before you start and arrange them in an orderly fashion.”
The first BSL-4 high-containment laboratories were built in the 1970s. Until then, researchers had handled extremely hazardous biological agents in so-called glove boxes: hermetically sealed, transparent cabinets fitted with rubber gloves, compatible with a class III BSC. To day, most BSL-4 high-containment laboratories operate as suit laboratories, where researchers wear full-body positive-pressure (“space”) suits (156; European Union Directives 90/218/EC, 98/81/EC, and 2000/54/EC). Key features of BSL-4 high-containment laboratories are the safeguards to prevent failure and faults of containment systems and measures. There is thus redundancy of critical systems and biosafety procedures.
Although the objectives that the legislators want to reach by using containment levels and procedures are not always explicitly mentioned, one could deduce some guiding objectives, as we have summarized in Table Table88.
The scientific literature on evaluation of the effectiveness of biosafety measures is very scarce and does not provide a consensus approach. The effectiveness of biosafety measures may be evaluated by different approaches and at different levels. In Table Table99 we have given a simple classification of approaches that collectively may provide guidance in evaluation activities. This table presents questions and purposes of single evaluation activities. A first level of evaluation may be directed at measuring the effectiveness of single containment equipment and procedures, such as the filtering capacity of face masks and BSCs under experimental circumstances. Such an evaluation could be directed at physical or, preferably, microbiological criteria. Subsequently these single apparatuses and procedures should be evaluated during practical work. It is evident that by taking this step from experimental challenge to practical work, unforeseen circumstances that may occur during practical work may be detected. For example, turbulences caused by movements of personnel during practical work may lower the protection afforded by BSCs. Air leakage may occur along respiratory masks during work. Masks may not fit perfectly.
A subsequent level of evaluation would be the laboratory as a whole, including its design and construction, the equipment, and working instructions. Again, such an evaluation can be done experimentally during a validation of the laboratory process or actually in a working laboratory setting. An experimental approach may use the deliberate release of indicator particles or model microorganisms. Evaluation of laboratory safety under field circumstances may include analysis of environmental samples taken inside and near the laboratory. In this case the effectiveness of biosafety measures and working instructions are actually evaluated during practical work and includes compliance of workers with working instructions, their experience and training, unintentional incidents, and efficacy of containment measures.
Finally, one may evaluate the effectiveness of measures at the clinical-epidemiological level, examining the overall effectiveness of measures in their capacity to prevent infection of laboratory workers and others. It is evident that laboratory workers play a central role in such an evaluation, as they are the persons who both are at high risk and may pass infections to others. Such epidemiological studies may follow a passive or an active searching approach. Clearly, for ethical reasons this level of evaluation is usually not suitable for an experimental approach.
We consider that optimizing the training of personnel and monitoring their compliance with procedures are important, as the best biosafety measures are only as good as the participation and discipline of laboratory workers themselves. In many reports, extensive training of laboratory workers and the proper execution of guidelines are mentioned as one of the most important measures to prevent incidents (85). However, poor compliance has also been reported (45, 144). In Argentina, incidents with pathogenic microorganisms were reduced after setting up a training program and providing protocols (85). The importance of training was also emphasized by experiences during the outbreak of severe acute respiratory syndrome (SARS) in 2003 (81, 100). Laboratory escapes of the virus from BSL-3 laboratories occurred in Singapore, Taiwan, and Beijing because of breaches in good laboratory practice rather than failure of the facilities. Extensive contamination occurred because gloves were inappropriately worn and contaminated surfaces were not disinfected (81).
Because most infections in the laboratory occur via aerosols and infected material and surfaces (123), equipment directed at minimizing airborne infections has received most attention. There are three classes of BSCs with different levels of protection, i.e., classes I, II, and III (152). In addition, within BSC II there are three subtypes. Class II BSCs consist of a chamber with a small open front in which an airflow is generated to prevent microorganisms from escaping the chamber. Laboratory workers sitting behind the cabinet insert their hands and arms into the chamber. All objects and the arms of the worker can disturb the airflow and cause the microorganisms to escape. Class III BSCs are basically similar, but the front is completely closed and can be accessed via attached gloves.
BSCs should meet legal standards, as, for example, defined by the European Union (EN12469). BSCs have been improved significantly during recent years due to this EN12469 standard, among others. Unfortunately, literature on the containment efficiency of class II BSCs is scarce and is addressed mainly in older publications, and such literature is virtually absent for class III BSCs.
In general BSCs provide a good level of protection when operated and maintained correctly (107). However, in several older studies, before the introduction of EN12469, it was shown that personnel working with open-front BSCs can still be exposed to infectious doses of microorganisms (7, 74, 112). Barbeito and Taylor (6) investigated the efficiency of containment of a BSC under three different closure conditions and different air velocities. In the cabinet, between 105 and 106 microorganisms per cubic foot were released in 5 min. When the glove panel was removed, a human infectious dose was released, and the number of microorganisms that escaped containment increased with decreased air velocity. Moreover, an increase in human activity in the cabinet resulted in increased numbers of microorganisms escaping the cabinet. When the glove panel was attached, no microorganisms could be detected outside the cabinet. A remarkable finding was that when the glove panel was installed without the gloves attached, no microorganisms escaped from the cabinet. Their main conclusions were that laboratory workers are protected from infectious microorganisms only when they use closed BSCs with high airflow velocities and that the effectiveness of BSCs is compromised by the activity of the workers. Macher and First (86) performed measurements on exposure of workers to bacterial spores using a class II BSC with an adjustable work opening. Aerosols of bacterial spores were created inside the cabinet, and the escaping spores were measured. The work opening height appeared to be a significant predictor of spore concentrations outside the BSC (86). Similarly, airflow velocity was negatively correlated with the concentration of escaping spores. Human activity in the cabinet, such as hands moving through the opening, also resulted in the escape of spores. Spore concentrations in the operator's breathing zone were about 24 times higher than acceptable levels. Surprisingly, working in the rear of the cabinet was less safe than working in the front, since the close proximity of the body to the cabinet influenced the airflow. Thus, for safe working conditions it is essential to limit the movement of arms and hands by arranging the equipment in the most practical way. Heidt (54) also tested the efficiency of a class II BSC. This author concluded that the cabinet provided sufficient protection, since microorganisms escaped only at the highest densities of the test aerosol created inside the cabinet. Since the number of bacteria detected was very low, this was considered to be acceptable. Osborne and coworkers (107) investigated a number of BSCs and calculated the operator protection factors (OPFs), as assessed by still and latterly limited “in-use” KI-Discus tests. The OPF is defined as the ratio of the exposure to airborne contamination generated on the open bench to the exposure resulting from the same dispersal of airborne contamination generated within the cabinet (66). Most BSCs had OPFs higher than 100,000, except when room pressure changed or when drafts occurred in the laboratory. The performance of class II BSCs was shown to be affected by the movements of the worker, and some movements reduced OPF results as found before by Macher and First (86). However, the levels of failure were marginal. The OPF tests revealed that a selected class II unit provided the same OPF as a class I unit when properly used.
Although the literature is not unequivocal, it appears that the use of BSCs decreases LAIs significantly (54, 86, 107, 119). However, in a recent publication, Rusnak et al. (119) examined illness surveillance data archived from the U.S. offensive biological warfare program (from 1943 to 1969) and concluded that BSCs and other measures failed to sufficiently prevent illness caused by agents with lower infective doses in a high-risk research setting.
Though required in some countries (including The Netherlands), cabinet performance is not generally assessed. However, on-site containment tests indicated that 37 class II BSCs (all with adequate type test certification and including 18 new installations) failed to meet the OPF requirements as defined in BS 7526. Thus, testing for containment using an OPF test appears to be essential both at commissioning and during routine maintenance (22, 107).
While formaldehyde gas has been used for over 100 years for decontamination, the efficacy of this process remains controversial (95). Moreover, because of its toxicity, the use of formaldehyde itself requires containment procedures (66). Formaldehyde decontamination of BSCs is usually validated using spore strips and culture. Therefore, poliovirus, Mycobacterium bovis strain BCG, or Bacillus spores have been used. Bacterial spores on stainless steel appear to be resistant to decontamination, and using bacterial spores to validate decontamination is too slow. Therefore, commercial biological indicator tests have been developed, which may be an aid in detecting incomplete decontamination. Difficulties in obtaining effective decontamination by using formaldehyde gas have been demonstrated. Factors contributing to the effectiveness of decontamination by formaldehyde include the formaldehyde level, the relative humidity, the temperature levels, and the medium to be contaminated. Locations beyond the exhaust filters of BSCs were the most difficult to decontaminate.
Modern cell sorting equipment has become an important tool in microbiological laboratories (77, 110). Because cell sorters lead to aerosol formation and are not easily accommodated by regular BSCs, this kind of apparatus could cause the operators to become contaminated (77, 110, 125). In addition, their high costs often prohibit their incorporation within BSL facilities. To solve this problem, Lennartz and coworkers (77) integrated a fluorescence-activated cell sorter (FACS) into a specially developed class II BSC. Biosafety was subsequently tested by using T4 bacteriophage aerosols and shown to be excellent. Bacteriophages were readily detected inside and outside when the airflow of the BSC was off, but when the BSC was turned on no bacteriophages could be detected outside (77). Many FACS protocols include inactivation steps, including the use of fixatives based on alcohols or formaldehyde. Some of these protocols have been evaluated for antimicrobial activity directed against specific pathogens, in particular HIV. Formaldehyde at concentrations of 0.5 to 2% is effective in inactivating HIV, but the ability of fixatives to inactivate other microorganisms in FACS equipment, including HBV, has not been demonstrated. In addition some protocols employ nonfixed cells. While analytic cytometers are engineered not to produce aerosols, jet-in-air cell sorters generate droplets and microdroplets that may be aerosolized. Recently high-speed cell sorting using high operating pressures with an increased potential for aerosol generation and an enhanced risk of sample splashes at the sample introduction port has become more prevalent. At the same time, instrument manufacturers have become more safety conscious and have developed novel devices for containment of aerosols and splashes, modified sample uptake ports on cell sorters, and installed mechanisms to stop sample flow in case of a nozzle clog to reduce operator risks (111, 125). A Vantage FACS was thus modified for safe use with potentially HIV-infected cells. Safety tests with bacteriophages were performed to evaluate the potential spread of biologically active material during cell sorting. The bacteriophage sorting showed that the biologically active material was confined to the sorting chamber. A failure mode simulating a nozzle blockage resulted in detectable droplets inside the sorting chamber, but no droplets could be detected when an additional air suction from the sorting chamber had been put on (133). While these observations may be reassuring, some recommendations regarding the use of FACS equipment are important (124-126). The recently published International Society for Analytical Cytology biosafety standard for sorting of unfixed cells states that droplet-based sorting of infectious or hazardous biological material requires a higher level of containment than the one recommended for the risk group classification of the pathogen (126). Training has to include performing aerosol containment testing of instruments to be used for biohazardous sorting. In addition, waste fluid has to be collected in 10% sodium hypochlorite, and fluid lines should be disinfected using a 1:10 dilution of 5.25% sodium hypochlorite. Notwithstanding their potential hazards, no documented disease transmission through the use of a cytometer has occurred (124, 125).
A few papers examined the efficacy of face respirators and surgical masks. For example, Balazy et al. (5) examined the performance of two types of N95 half-mask, filtering face piece respirators and two types of surgical masks. The collection efficiency of these respiratory protection devices was investigated using MS2 virus (a nonharmful simulant of several pathogens) in a particle size range of 10 to 80 nm. Penetration of virions through N95 respirators, which are certified by the National Institute for Occupational Safety and Health (NIOSH), can exceed an expected level of 5%. The tested surgical masks showed a much higher particle penetration of the MS2 virions: 20.5% and 84.5%.
The proper functioning of equipment should be evaluated not only in isolation but also in the context of the entire laboratory. In considering biosafety of laboratories, the proper functioning of autoclaves may be overlooked. Barbeito and Brookey (7) and Marshall et al. (87) emphasized the potential of autoclaves to release viable microorganisms into the atmosphere and the importance of proper sterilizer location, ventilation, containment of heavily contaminated loads, and adequate sterilizer maintenance.
One of the few studies to assess contamination of the laboratory environment with pathogens found in blood examined 800 environmental samples taken from 10 clinical and research laboratories working at the BSL-2 level at the NIH. Thirty-one samples from 11 work stations in three laboratories contained HBV surface antigen. Factors associated with environmental contamination included flawed laboratory techniques (mouth pipetting, splashing, placing pens in the mouth, improper use of equipment, and improper instrument design requiring external wash steps) (odds ratio [OR], 9.78; 95% confidence interval [CI], 1.46 to 65.49), high work loads (OR, 5.06; 95% CI, 0.8 to 31.96), and inappropriate behaviors (including not wearing gloves) (OR, 2.75; 95% CI, 0.44 to 17.4). Flow cytometry was identified as the technique with the most frequent occurrence of overt spills (38). Indeed, HBV infection was among the most commonly reported LAIs. Laboratory workers in urban medical centers may have been at almost three times the risk of acquiring HBV infection than other hospital employees due to exposure to patients' blood and at 7 to 10 times the risk than that of the general public (38).
Some evidence of the effectiveness of a BSL-3 laboratory environment may be derived from experiences with a specially designed BSL-3 laboratory for autopsies of patients with SARS (79). SARS coronavirus is highly infectious, and during the outbreak of SARS more than 30% of the approximately 8,000 infected persons were health care workers. The autopsy laboratory was established in Beijing Ditan Hospital (which was designated the SARS hospital during the outbreak of SARS in China) in May 2003. Remarkably, the efficiency of decontamination in this laboratory was evaluated by a sarin simulant test. A sarin simulant aerosol of 0.3-μm particles at 4 mg/liter was generated and spread by a special device in the contaminated area. Sarin could not be detected in either the semicontaminated area or the clean area, and particles of >0.3 μm in size were not detected in the exhaust air. Twenty-three pathologists and technicians participated in 16 complete autopsies that were performed on patients with clinically confirmed or suspected SARS, of which seven cases were later confirmed to be SARS infections. None of these personnel demonstrated any evidence of SARS infection.
The set of biocontainment measures that define BSL-4 biosafety is comfortably the most comprehensive and stringent, but each setting and laboratory design is unique and comparative data on their containment effectiveness are nonexistent. A problem in assessing the effectiveness of BSL-4 containment measures is that isolated containment measures are considered insufficient. Thus, individual components such as autoclaves, incinerators, chemical decontamination showers, gaseous decontamination systems, air ventilation systems, and HEPA filters can be tested for physical parameters during normal operation and under extreme conditions, but it remains unclear how closely the simulated test conditions resemble the real-life situation. The effectiveness of HEPA filters is typically validated using bacterial spore strips or particles.
The analysis of laboratory accidents may illustrate what can go wrong and point the way to improvements. Such accidents are one of the most relevant parameters to evaluate the overall effectiveness of integrated biosafety measures. However, the epidemiology of the incidence and severity of LAIs is largely unknown, as there are neither national surveillance and monitoring systems with complete coverage nor many systematic studies on their occurrence (60, 131). Denominator data that are necessary to calculate the actual incidence of LAIs are usually lacking. In addition, LAIs may be subclinical and may have an atypical incubation period and route of infection, and laboratory workers and directors may be reluctant to report them because of fear of reprisal and stigma (52, 131). Therefore, much information is obtained from anecdotal case reports and some retrospective questionnaires. Such case reports do not always report on possible failure of biosafety procedures or unintended accidents. While accidental parenteral inoculation of infectious material appears to be one of the leading causes of LAIs, most LAIs appear to occur even with the best safety precautions in place (116, 131). A summary of some recent LAIs is given in Table Table1010.
Most LAIs are caused by microorganisms that are very pathogenic or that need a very low infectious dose, including arboviruses, Venezuelan equine encephalitis virus, hantavirus, HBV, HCV, Brucella sp., Coxiella burnetii, Francisella tularensis, Mycobacterium tuberculosis, Salmonella spp., Shigella spp., Chlamydia psittaci, Blastomyces dermatitidis, Coccidioides immitis, Cryptosporidium spp., and organisms causing typhus, streptococcal infections, histoplasmosis, leptospirosis, tularemia, coccidiomycosis, and dermatomycosis (52, 116, 149). A direct link to accidents or exposure events, such as aspiration, injection, cut, spill, or bite, appears to be apparent in only a minority of the LAIs, while the majority are likely caused by undefined exposure to aerosols (52, 116, 149-151, 159). Indeed aerosols have been responsible for major outbreaks of LAIs caused by Brucella spp., Coxiella burnetti (Q fever), Chlamydia psittaci (psittacosis), and M. tuberculosis (89, 103, 129). The main hazards for inoculation (114-116, 129, 131) include (i) parenteral inoculation, (ii) inhalation of infectious aerosols, and less commonly, (iii) accidental oral ingestion and (iv) direct contact with mucous membranes or (broken) skin. Special hazards occur when working with infected animals. The presence of highly pathogenic microorganisms in unknown clinical samples likely explains the high incidence of tuberculosis among laboratory workers. In different studies the incidence of tuberculosis in laboratory personnel is estimated to be 3 to 100 times the frequency observed in the general population. The high infectivity of M. tuberculosis is related to its low infective dose (i.e., a 50% infective dose of <10 bacilli) (1, 117). Schellekens (123) calculated that 1 out of 100 to 1,000 laboratory workers per year are infected, but recent studies suggest that the rate of LAIs per person per year is decreasing (107, 123, 156).
Pike (116) tabulated the most common sources of LAIs from published literature and survey data. In the period from 1924 to 1977, there were 4,079 reported cases of LAIs with 168 casualties. In the subsequent period from 1980 to 1991, there were 375 reported cases with 5 casualties. At the time of Pike's survey, most LAIs (59%) occurred in research laboratories, compared with 17% in diagnostic laboratories. The highest mortality rate (7.8%) was associated with psittacosis. At that time, approximately 70% of LAIs resulted from work with the infectious agents (21%) or animals (17%), exposure to aerosols (13%), and accidents (18%). Less frequent sources of infection included clinical specimens (7%), autopsies (2%), and contaminated glassware (1%). Most causes of LAI were unknown (82%), and in only 18% of the reported cases could the cause be attributed to accidents, associated with the use of sharps such as needles (25%), injuries by glass (16%), splashes or spills (27%), mouth pipetting (13%), and bites by laboratory animals (14%). Many of the LAIs of unknown origin were likely caused by exposure to an infectious aerosol.
A recent survey of symptomatic and asymptomatic LAIs has been conducted by Harding and Byers (52), who reviewed 270 publications from 1979 to 2004, a period during which much has been done to improve laboratory safety while the work load in laboratories increased. A decrease in the number of LAIs would therefore be expected; however, knowledge on the total population at risk and the total number of infections would be needed. Harding and Byers (52) found a total of 1,448 cases and 36 deaths, 6 of which were aborted fetuses. The infections occurred in clinical, research, teaching, public health, and production facility laboratories, with clinical and research laboratories accounting for approximately 76%. In recent years more LAIs from clinical laboratories were reported, probably due to a more active employee health program, the absence of biosafety containment equipment in a number of clinical laboratories, or the fact that during the early stages of culture identification, personnel are working with unknowns and may not be using adequate containment procedures. Like earlier findings, the authors report that only a small proportion of the LAIs resulted from actual accidents. Most were acquired by simply working in the laboratory or by exposure to infected animals.
Sewell (131) concluded that adherence to the guidelines promulgated by the various regulatory agencies decreases the risk of occupational exposure to infectious agents. However, he also recommended additional studies to evaluate the effectiveness of other safety measures implemented or mandated in the laboratory. Interestingly, Sewell (131) describes personal risk factors of laboratory workers that are associated with accidental infections. Characteristics of persons who have few accidents include adherence to safety regulations, a respect for infectious agents, “defensive” work habits, and the ability to recognize a potentially hazardous situation. In contrast, persons involved in laboratory accidents tend to have low opinions of safety programs, to take excessive risks, to work too fast, and to be less aware of the infectious risks of the agents they are handling. Also, men and younger employees (17 to 24 years old) are involved in more accidents than women and older employees (45 to 64 years old).
While many reports emphasize the importance of personal protection, there are indications that extensive personal protection by use of double gloves, face masks, and protective clothing is not the sole solution, since such measures can reduce the dexterity of the laboratory worker, leading to increased accidents (122). This indicates that the use of sharps should be minimized when workers wear extensive personal protection.
Laboratory-acquired parasitic infections, from both protozoa and helminths, have been extensively reviewed by Herwaldt (56). Importantly, because protozoa, in contrast to most helminths, multiply in humans, even a small inoculum can cause illness. The author summarizes 199 case reports on laboratory and health care workers. The most frequently reported parasitic infections were caused by Trypanosoma cruzi, Toxoplasma gondii, Plasmodium spp., Leishmania spp., and Cryptosporidium parvum. Two cases (one of Chagas' disease and one of toxoplasmosis) were fatal. However, as with other infections, accurate counts of accidental exposures and infections and information on the risk per person-year are unavailable. Some of the laboratory-acquired parasitic infections were directly linked to accidents (e.g., a bite by an escaped infected mosquito) and poor laboratory practices, such as recapping a needle, removing a syringe from a needle, working barehanded, mouth pipetting, and working too fast. For 105 cases an accident or a likely route of exposure could be presumed; 47 (44.8%) of these included a percutaneous exposure via a sharp object. In other cases no apparent accidents were recognized or reported, suggesting that subtle exposures (e.g., contamination of unrecognized microabrasions and exposure through aerosolization or droplet spread) resulted in infection.
Walker and Campbell (146) did one of the few systematic but retrospective studies. They carried out a retrospective questionnaire survey of 397 responding United Kingdom laboratories covering 1994 and 1995. Approximately 75% of these were diagnostic laboratories, 14% were research laboratories, and 9% were teaching laboratories. Over 55,000 person-years of occupational exposure were covered, and only nine cases of LAI were identified, giving an overall infection incidence rate of 16.2/100,000 person-years, compared with 82.7 infections/100,000 person-years found in a similar survey covering 1988 and 1989 which was conducted by Grist and Emslie (48). This decline in incidence continues the trend previously reported for the period from 1970 to 1989. Infections were most common in females (in contrast to the findings reported by Sewell ), in relatively young staff, in microbiology laboratory workers, and in scientific/technical employees. Gastrointestinal infections predominated, particularly shigellosis, but few specific etiological factors relating to working practices were identified. These included a broken glass leading to a hand cut, a rat bite, and aerosol contamination. In most cases no clear accident was reported. Lack of experience was cited as a definite factor in two of the cases. Single cases of HCV, E. coli O157, and M. tuberculosis infection were identified, in addition to single cases of nonspecified septicemia and gastroenteritis. The absence of any cases of HBV infection, as in 1988 to 1989, reflects a sharp decline since 1970 and was attributed to increased awareness, better technique, and the availability of immunization. Furthermore, the absence of eye infections and the paucity of skin infections may indicate good technique and use of protective equipment. Despite the shortcomings of this study (retrospective study design, no reliable denominator, potential underreporting or underrecording, and no detection of asymptomatic infections), the authors concluded that the small number of cases identified indicates high standards of infection control, although they still recognized room for improvement. Finally, the study emphasized that the notification system in place in the United Kingdom to report LAIs is inadequate for the task of monitoring their true incidence in a comprehensive way, a conclusion that probably holds true for many other countries. For comprehensive monitoring of the incidence of LAIs, it is necessary to establish a routine, active surveillance program or prospective survey which has the support and commitment of the laboratories themselves.
Recently, a report was drafted on the biosafety status of clinical laboratories in Japan (160). Data were obtained from 431 hospitals and 301 institutions. The authors found 28 cases of possible laboratory-associated tuberculosis infection, of which 25 could be associated with the lack of BSCs, which are required for work with M. tuberculosis. Other risk factors were insufficiently skilled equipment operation and rupture accidents during centrifugation of blood. Within the last 5 years 1,534 events of self-inflicted needle punctures were recorded (160).
A retrospective survey of incidents occurring during biotechnological and clinical work in Flanders, Belgium, indicated that on average 13.6 incidents occurred per year among 7,302 laboratory workers. As a result, 69 persons (<1%) were exposed to biological agents, resulting in 2 LAIs, caused, respectively, by L. monocytogenes and Brucella melitensis. Most incidents occurred in clinical laboratories, likely caused by the higher number of working hours actually spent in clinical laboratories and the sometimes unknown nature of microorganisms. Handling of experimental animals and waste was considered risky. Most incidents were caused by human failure, including prick accidents, spilling, breaking, and maintenance work carried out in the laboratory (30).
Sejvar et al. (130) undertook a systematic, retrospective evaluation of the risk of meningococcal disease among clinical microbiologists and an assessment of the laboratory procedures that might predispose technicians to infection. Cases of suspected or proven laboratory-acquired meningococcal disease were identified by placing an information request on e-mail discussion groups of infectious disease, microbiology, and infection control professional organizations. Sixteen cases of probable laboratory-acquired meningococcal disease occurring worldwide between 1985 and 2001 were identified, including six U.S. cases between 1996 and 2000. Nine cases (56%) were serogroup B; seven (44%) were serogroup C. Eight cases (50%) were fatal. In 15 cases (94%), isolate manipulation was performed without respiratory protection. An average of three microbiologists are estimated to be exposed to the 3,000 meningococcal isolates seen in U.S. laboratories, yearly resulting in an attack rate of 13/100,000 microbiologists between 1996 and 2001, compared to 0.2/100,000 among U.S. adults in general. The case/fatality rate of 50% seen among survey cases is substantially higher than that observed among community-acquired cases, which may be explained by ascertainment bias due to underreporting of mild cases of disease. However, an alternative possibility is that clinical microbiologists routinely work with highly virulent strains and high concentrations of organisms. All cases identified in this inquiry occurred among microbiologists and not among workers in other areas of the clinical laboratory. This suggests that exposure to isolates of Neisseria meningitidis, and not patient samples, represents the increased risk for infection. In addition, all isolates were derived from sterile sites. None of the microbiologists identified were working with isolates obtained from pharyngeal or respiratory secretions, suggesting that such pharyngeal isolates represent a lower risk, presumably due to their lower pathogenicity. The authors concluded that prevention should focus on the implementation of class II BSCs or additional respiratory protection during manipulation of suspected meningococcal isolates. Following two cases that prompted this survey, CDC has instituted a prospective surveillance for laboratory-acquired meningococcal disease (21).
In The Netherlands, two surveillance systems monitor the occurrence of labor-acquired infections, i.e., not exclusively LAIs. The number of reported LAIs is low, but both systems suffer from serious underreporting and do not provide details of the transmission route or accidental cause of infection (55).
Fortunately, the number of accidental releases or LAIs with GMOs appears to be very low. Jones et al. (59) reported the first case of accidental vaccination with a recombinant vaccinia virus (Western Reserve [WR] strain) expressing the nucleoprotein gene of vesicular stomatitis virus. The infection had a relatively mild course, perhaps because the laboratory worker had received smallpox vaccination 30 years before the accident or because of attenuation of the virus by the insertional inactivation of the thymidine kinase gene. Openshaw et al. (105) reported an accidental infection of a laboratory worker with recombinant vaccinia virus (WR strain) expressing proteins of respiratory syncytial virus. The infection occurred through two separate needle accidents during the same work session, although the worker was experienced in the procedure. The procedure was subsequently modified to prevent further accidents. The laboratory worker had been vaccinated with standard smallpox vaccine, a practice that may have restricted the severity of symptoms to local redness and swelling. Mempel et al. (89) reported the case of a recombinant vaccinia virus infection in a previously vaccinated researcher working with various genetically modified strains. The isolated virus carried a functionally inactivated cytohesin-1 gene of human origin, which impairs leukocyte adhesion by interacting with the LFA/ICAM-1 axis. The immunomodulating nature of the inserted construct might have added to the infectivity of the virus. Although the paper does not detail safety procedures in the lab, the infection occurred while the handling was considered proper. Contact infections were not reported. Lewis et al. (78) reported a case of ocular vaccinia infection in an unvaccinated laboratory worker. The infecting virus was a unique form of recombinant WR vaccinia virus constructed in the laboratory. Although laboratory staff generally followed established biosafety precautions, several opportunities for virus exposure were identified. Experiments were performed partly outside a BSC. Staff infrequently wore eye protection. Laboratory coat sleeves were not elasticized and did not always cover the wrist. Waste pipettes were not disinfected before removal from the BSC. Instances occurred in which samples with low titers of live virus were removed from the BSC, transported to other parts of the facility, and manipulated. In addition, laboratory staff routinely vortexed tubes containing live virus outside the BSC. Finally, a laboratory-acquired recombinant vaccinia virus infection running a severe course was reported in a research laboratory technician who had a long history of eczema. She had been working with a thymidine kinase-deficient strain of vaccinia virus for use as a vector for gene therapy. She did not report an accidental inoculation, but she usually wore no protective gloves. She had been vaccinated with smallpox vaccine as a child but not again before beginning her laboratory work with vaccinia virus (83).
Another case of laboratory-acquired vaccinia virus (nonrecombinant) infection was reported by Wlodaver et al. (158). This infection occurred in a laboratory technician who had not been previously vaccinated and who developed generalized vaccinia. She had accidentally cut a finger on a coverslip while working with vaccinia virus. Evaluation of this accident in her laboratory prompted a review of procedures for handling contaminated glassware. Moussatché et al. (94) reported another accidental needle stick inoculation of a laboratory worker with vaccinia virus. Although the patient had previously been vaccinated against smallpox, severe lesions appeared on the fingers.
In total, there have been at least 19 reported cases of laboratory-acquired vaccinia virus infections, of which 5 were accidental infections with recombinant vaccinia (83; this paper). Several researchers emphasize the necessity of vaccinia vaccination of laboratory workers. In the United States and Canada, specific recommendations exist for laboratory personnel who conduct research with (recombinant) orthopoxviruses, including vaccinia virus (20, 153). The Advisory Committee on Immunization Practices recommends revaccination at least every 10 years for persons working with non-highly attenuated vaccinia viruses, recombinant viruses developed from non-highly attenuated vaccinia viruses, or other nonvariola orthopoxviruses. To ensure an increased level of protection against more virulent nonvariola orthopoxviruses (e.g., monkeypox virus), empirical revaccination every 3 years can be considered. In contrast, mandatory guidelines with respect to vaccinia vaccination do not exist in Europe (57).
In conclusion, LAIs with GMOs appear very seldom and appear to be restricted to infections with recombinant vaccinia virus. Although laboratory accidents with other GMOs may have been unnoticed due to a subclinical course of infection, this situation seems to reflect that vaccinia virus is very widely used. Perhaps more important, the recombinant virus is still pathogenic, and this might be enhanced by certain gene inserts. Vaccinia virus infection can be established via several routes, including breaks in the skin, and the infectious dose is probably low. Guidelines for working safely with vaccinia virus, which include vaccination, are available (57). We consider it nonetheless advisable to work with the highly attenuated strains of vaccinia virus (modified vaccinia virus Ankara and NYVAC) or with avian poxviruses that have a restricted host range and do not replicate in mammals (ALVAC and TROVAC) whenever possible.
LAIs with category 4 biological agents (filoviruses, arenaviruses, flaviviruses, and bunyaviruses) are extremely rare and usually occurred earlier in settings with lower levels of biocontainment and/or involved animal work (41, 42, 51, 91, 127, 138, 139, 114). Rare laboratory incidents with New World arenaviruses have been reported in earlier surveys more than 4 to 5 decades ago (e.g., with Junin virus and Machupo virus) (51). Such experiences illustrated the need for more effective measures to reduce hazards.
While this low number of BSL-4 laboratory accidents may be reassuring, the number of BSL-4 labs and workers is increasing. This appears to be in defiance of the “concentration and enclosure” principle, because the risks associated with this work may increase with the number of facilities and workers (61, 62). Concern is further fuelled by several incidents that included unreported infections (among others involving Brucella and Coxiella burnetti) and other biosafety breaches. In 2006, a Department of Health and Human Services Inspector General audit of security procedures found that 11 of 15 institutions had “serious weaknesses,” such as unlocked doors and freezers and lax inventory records (61, 62). Another incident, in 2007, was the escape of foot-and-mouth disease virus from the Pirbright facilities in the United Kingdom, which has been linked to an outdated effluent system and caused several outbreaks of this very contagious disease among cattle and sheep (53).
In this paper we have reviewed the principles underlying biosafety measures for work with pathogenic microorganisms and GMOs, and we have examined to what extent evidence for their effectiveness is available. Clearly the risks of working with GMOs are considered to be largely identical to those of working with pathogenic non-GMO microorganisms, and hence much of the knowledge of and containment measures for GMOs are derived from the latter. Regulations appear to be more strict for GMOs, however.
Among many reports on biosafety, we found only scarce information on the evaluation of effectiveness and on criteria to judge effectiveness. We must therefore keep in mind that safety cannot be expressed in absolute terms. It is a relative concept defined in terms of tolerability and acceptability limits (64). This notion implies that workers and regulators try to find a balance between the costs of safety measures and the potential benefits of the work for society. For example, in microbiological work, safety measures and associated costs increase from BSL-1 to BSL-4. Indeed, safety measures at BSL-1 and -2 are probably insufficient to prevent all infections with microorganisms of the corresponding risk categories, but their consequences at these levels are considered to be acceptable or negligible.
The current biosafety practice gradually evolved during the previous century. Therefore, it is not immediately obvious whether and what principles have been employed to ensure safe work and on which scientific basis they were built. In this paper we have tried to identify some principles that appear to underlie the current practice. Such principles clearly partly overlap and mutually enhance each other. A central activity, either implicit but preferably explicit, is a thorough risk assessment procedure that considers all potentially harmful effects and their possibility of occurrence. Other important underlying principles are the use of (wherever possible and appropriate) biological containment, concentration and enclosure, exposure minimization, physical containment, and hazard minimization.
Clearly, throughout the world regulators have adapted the model of universal precautions based on a classification of microorganisms in four hazard classes and accompanying standard safety practices (16, 156). The advantage of this model is that work with certain microorganisms can be grouped together to comply with the accompanying containment rules according to their classification. A disadvantage may be that this universal model may overlook the necessity to tailor safety measures for specific microorganisms or specific strains with particular routes of transmission or virulence properties. Researchers should also reckon with variability in human immunocompetence. Therefore, risk assessment remains at the core of any individual experiment. Such a risk assessment should in particular be based on (preferably quantitative) parameters of transmission, infectivity, and virulence. These should guide the subsequent measures aimed at reducing the amount of microorganisms to which individuals are exposed to below a minimal threshold level of infectivity. Nonetheless, this universal model of four biohazard classes appears to work well, but we recommend further harmonization of criteria for both non-GMO and GMOs and between different regulatory authorities, such as the European Union, WHO, and CDC.
Altogether, the regulations specifying the biosafety containment measures appear to be based on experience, expert judgment, and common sense. They are not motivated or supported (at least not explicitly) by scientific literature, however, and often are not based on precisely defined or specified properties of microorganisms and vector and insert sequences. In addition, the regulations do not exactly specify the level of protection that they aim to afford, for example, in terms of diminishing exposure of the laboratory workers below a threshold level of infectivity. Furthermore, it is clear that the physical containment classes 1 to 4 afford increasing levels of containment, but it is not sufficiently clear and scientifically supported to what extent they provide effective protection with regard to prevention of infection of laboratory personnel, prevention of airborne escape, etc. This, together with sometimes not very detailed regulations, puts much responsibility on researchers, lab directors, advisors, and regulatory authorities in further detailing working practices. The regulations also do not comprise evaluation procedures to monitor the compliance or effectiveness of the containment provisions. Table Table1111 summarizes our recommendations.
The hazard classification of work with GMOs follows the classification of work with non-GMOs. This extrapolation should be based on a risk estimation as precise as possible considering the genetic modifications involved. In case of doubt or uncertainty about the properties of the GMO involved, regulators and biosafety experts will choose a higher risk classification than the risk level of the microorganism from which the GMO has been derived, or they will demand additional safeguards. Basic research on transmission properties of GMOs, in comparison with those of the nonmodified organisms, may be helpful in such a risk assessment to further define the risks involved in the manipulation of GMOs. However, often properties, such as infective dose, may be difficult to obtain. In risk assessment we consider it important to take gene-gene and gene-environment interactions into account, because experience (for example, with the IL-4/ectromelia construct) has shown that specific gene products may have unwanted effects in a particular environment (58).
Regulators do not always require routine evaluation and monitoring of biosafety aspects in laboratories, which we would like to recommend. Routine monitoring of biosafety aspects, including monitoring of compliance and educational and behavioral aspects, may not easily be implemented, in particular in the many clinical laboratories with their high workload involving a wide variety of sometimes unknown microorganisms, but it may enhance overall safety awareness. For example, validation experiments using T4 bacteriophage, bacterial spores, or other indicator microorganisms could be useful for this purpose.
From the literature it appears obvious that there is little experience and no consensus on how the effectiveness of biosafety practices should be evaluated. Clearly, the effectiveness of biosafety measures can be assessed at different levels and under different circumstances that logically complement each other; i.e., one could question whether a single piece of equipment is effective under experimental conditions or, conversely, whether the population has not been accidentally exposed to LAIs. Data on the biological containment efficiency of equipment and laboratories are scarce and fragmented and are mainly limited to technical specifications. Monitoring of LAIs therefore appears to play a pivotal role in evaluating the effectiveness of containment and the potential exposure of laboratory workers and the population, but it suffers from serious underreporting throughout the world. Many reports of laboratory accidents are only anecdotal. We therefore recommend optimization of the systematic monitoring of laboratory accidents, including the serological monitoring of personnel. Infection with microorganisms, either GMO or not, that have a high infective dose or low virulence, usually belonging to risk category 1 and 2, may be difficult to detect. The extent of serological monitoring should therefore depend on the risks involved. A passive sampling strategy, i.e., collecting serum samples at the time of employment and following incidents, may be sufficient for work with low-virulence microorganisms, but an active sampling strategy at regular intervals may be considered for class 3 and 4 microorganisms. One clue to optimizing monitoring of accidents may be the introduction of “blame-free” reporting, which aims for workers to share experiences without being punished. In addition to systematic monitoring, retrospective surveys may be very useful, as they may identify certain risk factors, as shown for the occurrence of meningococcal disease (130).
Despite the methodological imperfections, it is clear that the number of GMO-associated laboratory accidents is very scarce in comparison with the number of non-GMO-associated infections and is practically restricted to accidental infection with recombinant vaccinia virus. We interpret this finding to mean that the biological containment obtained by attenuating GMOs is possibly a major factor in preventing their transmission. However, other factors contributing to the low number of GMO-associated accidents may be that GMOs are well characterized, implying that the worker has knowledge of the properties of the GMO and that the work load involving GMOs is likely much lower than that in clinical laboratories. Moreover, clinical samples may contain unknown pathogens. Other factors that may contribute to this low number of accidents involving GMOs are the stricter regulatory framework and a stricter compliance with containment rules. In many countries, including The Netherlands, both the researcher and the regulator make a risk assessment for each individual project that involves GMOs, a practice that is less developed for work with non-GMOs. In case of doubt or uncertainties regarding the properties of GMOs, biosafety experts and regulators will demand a higher risk category or additional measures. Both a local biosafety officer and a national inspectorate supervise this practice. Whenever possible, we consider it important to further optimize the possibilities of employing genetic modification to enhance the safety of GMOs. In particular, we recommend further definition of the genetic properties underlying the transmissibility and infectivity of microorganisms, measurement of the influence of specific mutations on infective dose and transmission properties of GMOs, and that containment rules be based on such findings. This is not an easy task but would provide a further scientific basis for the phenotypic properties of GMOs and the accompanying level of biological containment afforded by specific genetic alterations. Vaccinia virus or recombinant viruses developed from non-highly attenuated vaccinia viruses appear to be less well suited as vector organisms due to their retained virulence and low infectious dose and should be replaced by safer poxvirus vectors wherever possible.
In many reports of LAIs there has been a lack of compliance with biosafety practices. This observation may be reassuring regarding the effectiveness of such biosafety practices, at least if they are followed. It illustrates that education of laboratory personnel and compliance with the rules remain the top priority. Increased attention to these aspects may have caused a decrease in the rate of LAIs per person per year (107, 156). On the other hand, in the majority of cases of LAIs a direct cause could not be assigned (52, 116, 131, 149, 159), suggesting that a failure was not noticed in many cases or that containment may have been insufficient. This observation may warrant further research on the routes of exposure in such cases and on the effectiveness of measures. Finally, although monitoring of LAIs is an important element in evaluating the effectiveness of containment measures, it may overlook the risks associated with nonreplicating agents, such as transduction by nonreplicating viruses.
Many countries, including The Netherlands, regulate work with pathogenic microorganisms and GMOs differently. Because the regulations are derived from the same underlying principles and use the same instruments for biosafety and because the number of accidents involving GMOs is very low, we recommend harmonization, modernization, and simplification of the regulatory framework through developing a single set of regulations for both non-GMOs and GMOs.
Despite their presumed overall effectiveness in providing biosafety, it is often unclear to what extent the current set of specific biological or physical containment measures, alone or together, contribute to the prevention of transmission of pathogenic microorganisms or GMOs. In further developing and modernizing the biosafety practice, we therefore recommend developing evidence-based practices and criteria to evaluate effectiveness wherever this is possible and feasible. This may optimize and perhaps simplify future biosafety measures and stimulate compliance with the rules. Although scientific research may strengthen the evidence base for biosafety measures, such work is complicated and does not necessarily guarantee new findings on which further improvements can be based. To unravel complexities and to obtain further insight into the contribution of specific elements to biosafety, mathematical modeling, which is directed at quantitative parameters of infectivity and transmission, may be supportive, but modeling obviously needs confirmation by observational and experimental findings. Such an approach may, however, point to the data that are needed to further guide the development of evidence-based risk analysis and containment policy for both non-GMO pathogens and GMOs.
This work was financially supported by the Dutch Committee for Genetic Modification (COGEM).
We thank Marja Agterberg, Marjolein van Esschoten, Ben Peeters, Erik Schagen, Gijsbert van Willigen, and Dick van Zaane for their constructive support.