Memories of a virulent past

Abstract

Studies on bacterial virulence and host-pathogen interactions usually rely on type strains isolated from human or veterinary populations. For instance, the El Tor strain N16961 has been extensively used to characterize Vibrio cholerae virulence, while E2348/69 is a default choice for studies on enteropathogenic E. coli interactions with host cells. Subsequent to isolation, such strains are passaged under laboratory conditions in rich medium, and often genetically manipulated to induce specific mutations or other alterations. While the cumulative knowledge gained by focused studies on a limited number of bacterial isolates allows for rapid scientific progress, strain diversity resulting from prolonged propagation and manipulation in different laboratories may lead to a distorted perspective and, at times, confound attempts to replicate specific experiments. Exploring such aberrations is an inevitable and necessary, if not always welcome, part of scientific progress.

Keywords: :

• reproducibility
• replication
• cell lines
• strains
• scientific method

Science is messy, and scientific progress seldom follows a straight line. First, it should be no surprise that scientists are subject to the motivations and foibles typical of all human beings (and they should be held accountable for such behavior). Beyond that, the tools and methods used to address scientific questions often have such serious limitations that it is a wonder that we get anything right. In a thought-provoking article on the reproducibility of scientific results, cancer biologist Mina Bissell expressed concern about the drive from the NIH, and some journals, to put pressure on investigators to demonstrate that their experiments can be replicated.¹ While Bissell clearly supports the necessity for replication in drawing meaningful scientific conclusions, she articulated several difficulties in achieving it. Her first concern related to the authority and expertise of external “validators,” and whether the costs imposed on individual scientists even made the exercise practicable. Her second point was that the ability to replicate experiments required a high level of training and skill; not all experiments could be readily reproduced by reading the “Methods” section of a publication.

Bissell’s article engendered a robust and varied response from the research community, including this one: “A result that is not able to be independently reproduced, that cannot be translated to another lab using what most would regard as standard laboratory procedures is not a result. It is simply a ‘scientific allegation’.” A more recent newspaper article on reproducibility, which did not directly address the article, used a more colorful turn of phrase to express the same sentiment: “If a result appears only under the full moon with Venus in retrograde, is it truly an advance in human knowledge?”²

Concerns about logistics and cost are perhaps not material to a philosophical discussion on how research should be performed, while the problem of “sensitive” experiments is an admission that not all parameters in an experiment may be known with adequate precision to be recorded in the “Methods” sections (water and air quality, and microbiota differences are some influences that come to mind). I remember constantly worrying about such issues in graduate school, and wondering if my experimental data had any meaning in the “real” world. Not arriving at a satisfactory answer, I resorted to titling my laboratory notebooks “Artifacts I,” “Artifacts II,” and so on, leaving my research advisor considerably perplexed and disconcerted.

Bissell provides some examples that likely resonate well enough with meticulous scientists practicing their trade with the utmost integrity and thoughtfulness. She indicates the exquisite “sensitivity” of epithelial cell lines, noting that minute alterations in their growth conditions could trigger epigenetic and genetic changes, and result in distinct phenotypes. This should be quite familiar to researchers working on intestinal physiology. To start with, immortalized intestinal cells are such distant approximations of normal tissue that their continued elaboration of intestinal phenotypes should be something of a surprise. No self-respecting normal intestinal epithelial cell would even brook the ploidy variations of immortalized cells: Caco-2s have 96 chromosomes, T84s have 56, and HT-29s have 71. These cells, as well as sub-populations with properties distinct from the parent line (e.g., Caco-2_BBE), have been transmitted and propagated in different laboratories.³ Any laboratory that has borrowed T84 cells from another (or purchased it from ATCC), even after karyotype validation, will attest that the cell lines are clearly different, with differences ranging from growth characteristics to cytokine production.

In vitro culture conditions are so far removed from growth in the intestine that temperature is likely the only parameter even approximating in vivo conditions. The medium is artificial, microbiota and immune cells are absent, and the environment is relatively static. To paraphrase an anoxia researcher: “You would not think of doing experiments routinely at 20 °C; how is it that you use cells grown in an oxygen-replete incubator (with 5% CO₂, instead of colonic conditions with limited, and variable, oxygen levels) for all your studies?”

Host-pathogen interactions using such cultured cells bring in the added complexity of bacterial strain variations resulting from laboratory passage.³ Many archetypal strains routinely used for such studies are initially isolated from patients, but have since been cultured in rich media at 37 °C, or frozen away in glycerol at -80 °C, neither condition resembling anything encountered by an enteric pathogen in “real” life. After repeated culture in disparate laboratories, they still retain their names, but are likely very distant and different relatives of their pathogenic forbears.⁴ The pathogenic properties they do retain are but memories of a virulent past. Two striking examples illustrate this point.

The enteropathogenic Escherichia coli (EPEC) strain E2348/69 was isolated over 45 yearrs ago from a child in the UK during an outbreak of infantile diarrhea.⁵ A minimally passaged isolate of this strain, sequenced in 2008, contains the pE2348/69 plasmid which confers streptomycin resistance (Str^R). E2348/69 has been widely employed to study EPEC pathogenesis, but the strain used in most laboratories is streptomycin-sensitive and nalidixic acid resistant. Streptomycin sensitivity (Str^S) resulted from loss of the pE2348/69 plasmid, while nalidixic acid resistance (Nal^R) was intentionally selected for, in order to facilitate recovery in volunteer studies. The “laboratory” strain has a mutation in gyrA, shown previously to confer Nal^R.

In addition, this strain has two other nonsynonymous chromosomal mutations compared with the Str^R parent strain. A mutation in ftsK, predicted to result in a conservative amino acid change, did not appear to significantly alter protein function. The other mutation altered a stop codon at the end of hflD (involved in inhibiting lambda lysogenization), fusing it to downstream purB, a purine biosynthetic gene. Consistent with a compromised function for the fused PurB protein, the Nal^R laboratory strain grew slower than the Str^R strain in minimal medium; adenine supplementation, or complementation with plasmid-encoded purB considerably ameliorated the growth defect.

The implications of this partial purine auxotrophy to virulence may be considerable. The Nal^R derivative is ten times less efficient at invading epithelial cells compared with the parent Str^R strain, and this defect is eliminated by purine supplementation. Although EPEC has been observed within epithelial cells in small-intestine biopsy specimens from infected children, it is primarily considered to be an extracellular pathogen. It is possible, therefore, that the atypical behavior of the lab-adapted strain has led to our diminished appreciation for EPECs intracellular lifestyle.

In a similar vein, two high-profile studies on the role of the Clostridium difficile toxins, TcdA and TcdB, in virulence arrived at somewhat different conclusions.^6,7 While the study by Lyras et al. concluded that only TcdB was essential for virulence, Kuehne et al. reported that the presence of either toxin was sufficient to cause disease. The integrity or competence of the two teams was never in question, and the methodologies and conclusions of the two papers were not immediately disputable. More recent work suggests that the differences could be, at least partially, explained by the C. difficile strains used. Both groups started with “strain 630” (originally isolated from a patient in Switzerland in the 1980s), but had independently passaged it to generate erythromycin-sensitive derivatives that could be genetically manipulated (JIR8094 and 630ΔErm, respectively).

Reproducibility is, and rightly so, the gold standard in science. Nevertheless, these striking examples indicate that it is not easy to achieve, and often, even by scientists with the highest integrity and talent. Still, it is worth remembering that a lack of reproducibility typically leads to unanticipated insights and new lines of investigation. Recent work exploring the differences between the two C. difficile strains revealed that 630ΔErm produced three times more toxin overall, expressed significantly greater levels of flagellar genes, and displayed robust motility compared with JIR8094.⁸ These findings suggest that the roles of TcdA or TcdB have to be considered in the context of the strain background and the contributions of other virulence attributes. This mirrors the overall trend in C. difficile field, where the critical roles of non-toxin virulence determinants are being increasingly appreciated.

One may aver that the inability to replicate experiments, due to experimental conditions, or strain and tissue culture variations, slows down scientific progress, and wastes taxpayer money. Such purists are likely far removed from the bench and the realities of the avocation. For instance, bacterial strain variation is inescapable, not just in the laboratory, but even within individual patients during the course of an infection.⁹ More broadly, for much of human history, gross misunderstandings of the natural world gripped our collective minds (and perhaps continue to do so), before they were reluctantly shed in the light of mounting evidence. Irreproducible experiments offer us an opportunity to set the record straight, acknowledge the inevitable truism that all model systems are only approximations of “reality,” and help us refine our understanding and improve our methods.

Disclosure of Potential Conflicts of Interest

No potential conflict of interest was disclosed.

10.4161/gmic.28340