Chapter I

Which digression on biogenesis battles (the sort of thing that would have been slipped into a lengthy footnote in the old days) returns us at last to the main path, to explore what was going on down inside those “mind boggling” bacteria, and see whether Michael Behe’s cozy notions about the “limits of Darwinian evolution” squared with the details of the metallo-β-lactamases.

As seen through the lens of modern evolutionary phylogeny (a term coined by neologism lover Haeckel), where geochronology cannot be avoided in the way Pasteur appeared to have done in the 19th century, the MBL enzymes are an extremely ancient group in bacterial biology, the B3 family tracking back several billion years, with the distinct B1+B2 cluster originating a billion years later, as explored by Barry Hall et al. (2003; 2004)—that would be the Barry Hall of the paper Behe was using at Dover. The β-lactamase module shows up in a broad range of eubacterial functions, including RNA/DNA processing and repair, filtering in turn down the evolutionary pipeline all the way to insects and mammals, Daiyasu et al. (2001).

Concerning the modern players causing consternation among medical practitioners, the relevant feature of IMP-1 appears to have developed from IMP-3 (from which it differs by only two amino acids, at positions 126 and 262) by a single point mutation (guanine to adenine) at the codon level for position 262, Iyobe et al. (2000) and Materon & Palzkill (2001, 256). The replacement of IMP-3’s glycine with serine expanded IMP-1’s contact range for the hydrolysis of a variety of antibacterial drugs (including imipenem) that was the reason for Hall’s 2004 caution flag.

Flip IMP-1’s position 262 back to glycine (again by single point mutation) and you get IMP-6, Oelschlaeger et al. (2005), again with altered potency against many of the carbapenem antibiotics (a battery of ten would be in use by 2012) but fortunately at a tradeoff of lowered resistance to the venerable penicillin and the newer imipenem. And then there’s IMP-25, E. Liu et al. (2012), which varied not only at S262G but differed from both IMP-1 and IMP-6 at G235S. That second minor mutation represented yet another conformation shape change from its cousins, leading to still more resistance to some of the carbapenems but less against others due to how its active edges interacted with neighboring molecules. Neither of those mutations were able to overcome the effectiveness of imipenem, though, confirming Hall’s 2004 prediction about the relative security of that antibiotic when dealing with forms close to the IMP-1 model.

But why was that so? Behe never thought to ask, an example of ID ball dropping that becomes particularly interesting once all the details are put on the table.

As researchers followed up with the clinical investigation Hall recommended back in 2004, it turned out that the IMP family was very busy indeed, and not alone. Thirty-three distinct IMP strains had been uncovered by 2012, along with the same number for the related VIM family, and half a dozen more in a newly discovered NDM enzyme group, Widmann et al. (2012) and Meini et al. (2014). Some were single residue variants like IMP-6, IMP-10 and IMP30, while others had many more (IMP-18 differed from IMP-1 in 48 places) with similar variety for the VIM family. Mutation hotspots “unexpectedly distant from the active site” were also discovered around the “ββ sandwich” part of the protein (at 208/266 for the IMP family and 215/258 for the VIM family) that turned out to be involved in the hydrogen bonding network that ultimately affected their catalytic efficiency (only asparagine or serine would do as a change at 208 in IMP, for instance, but once in place it was conserved in subsequent selective evolution).

A recent review of MLB evolution (natural and experimental) illustrated this “more than one way to skin a cat” approach to how this natural system had accomplished its antibiotic resistance:

A comparison of the available mutants in the numerous VIM and IMP families of MBLs surprisingly reveals that these groups have evolved their substrate specificities by different mechanisms. VIM enzymes accumulate mutations in the loops flanking the active site, mostly affecting the electrostatic interaction with the substrate carboxylate or loop flexibility. The same mechanism is observed for some IMP variants. In addition, substrate preferences among IMP variants are tuned by second sphere mutations. Meini et al. (2014, 307).

Similar diversity has appeared concerning the third player in this deadly game, the NDM family of MLBs, D. King et al. (2012), T. Li et al. (2013), and Zheng & Xu (2013), with NDM-1 transpiring as “the ultimate promiscuous enzyme,” Y. Kim et al. (2013). A fungal compound has been found that blunts the resistance effects of NDM-1 (for the time being), Meziane-Cherif & Courvalin (2014) re Andrew King et al. (2014), with apropos perspective by Ann Reid (2014a-b) for the NCSE’s Science League of America on how the evolutionary dynamics of natural mutation and bacterial plasmid exchanges are only slowed by such efforts, not banished. The ability of the component parts to sport all manner of resistance potential is highlighted further by Song & Tezcan (2014), artificially engineering new MBL forms to explore their properties.

All this means that the proteins (which, for old-timers like myself who grew up with land lines, resemble in schematic form—as most proteins do—a blobby tangle of linked phone cord coils) could naturally vary substantially in size and shape based on which mutations occurred and where, leading to different interactions in the bacterial environment. Position 235 lies on the IMP enzyme’s surface, for example, and directly affects protein recognition when glycine is replaced by serine, while position 67 is at the base of the β hairpin loop whose side chain leads to the active substrate recognition site, and the presence of any of four particular amino acids there (cysteine, serine, threonine or tyrosine) evidently contributed to the slightly increased resistance to imipenem noted by Hall (2004a).

But a far more salient point concerns whether a feature of the bacteria strain Hall used for his study (Escherichia coli) itself skewed the outcome. E. coli had been an obvious choice because it was both common and a prominent player in the field—representing 88% of the culprits in a Bangladesh study of urinary tract infections, for example, Lina et al. (2007). But its cell membrane happens to be especially permeable to imipenem, Matsumura et al. (1999), meaning that particular antibacterial agent was usually able to slip through the outer membrane faster than any of its competitors, evading by sheer molecular numbers the scattershot interference of the MBLs that might bind to it in the cellular fog.

The upper “limit” to the IMP-1 example might then be more a correlate of that diffusion rate, not the ability of IMP-1 to mutate substrate specificity to it (the evolutionary brick wall Behe wanted it to be). In that case, Hall’s prediction about the prospects for increased sensitivity to imipenem in IMP1 based on his assayed reactions wouldn’t have been because “Darwinian evolution” had struck the sort of limit Behe imagined, but because the appropriate mutations had already taken place but were unable to translate into any further resistance because in the leaky environment of E. coli it couldn’t get any more bang out of its already full MBL arsenal.

So what was Nature up to in this realm of Darwinian limits? Well, the relevant bacteria seemed not to be reading The Edge of Evolution. At the same time Behe was name-dropping Barry Hall, Walsh et al. (2005) were warning how “nearly 30% of imipenem-resistant Pseudomonas aeruginosa strains possess a metallo-β-lactamase.” (A further irony here is that Mark Toleman was among the co-authors of the Walsh paper.) The virulence of P. aeruginosa was no surprise: it was a veritable integron magnet, surveyed by Strateva & Yordanov (2009), by which time it was accounting for 15% of hospital acquired infections worldwide. But to our point, the MLB in question was none other than plain old unmodified IMP-1. First detected in Japanese strains of P. aeruginosa in the 1990s, by 2004 (and hence knowable in principle to Behe in 2005) it was showing up in Singapore and China, Koh et al. (2004) and C. Wang & Mi (2004), often with silent mutation allele variations reminding us of the ever-playing mutational baseline running under the protein assembly dance.

And consider further that P. aeruginosa is one of those bacteria with a low membrane permeability, C. Wang & Mi (2006) and Lister et al. (2009). The busy OprD porin channels found in E. coli that offer imipenem easy entry are not expressed in P. aeruginosa. Instead, it relies on a much more restrictive water-filled porin system (with an analog also known in E. coli). Only 8% as permeable as the OprD method, it is just sufficient to let vital nutrients in—and judging by some European strains, P. aeruginosa may be able to manage the porin gateways so well that imipenem has no chance even without IMP or VIM genes, Strateva et al. (2007) and Sacha et al. (2012).

Overall, imipenem continues to be effective against the membrane-leaky E. coli vector, such as urinary infections in Iran, Falakian et al. (2011), though resistance to imipenem was found in a European E. coli strain after “horizontal gene transfer and chromosomal mutations” affected expression of its membrane system proteins, Poirel et al. (2004). Meanwhile, a vector sensitive susceptibility to imipenem has cropped up experimentally concerning IMP variants in P. aeruginosa: Toleman et al. (2003) again on the transposon hitchhiker IMP-13 and Jeannot et al. (2012) on IMP-29, where lowered resistance to imipenem reappeared when they were transferred for evaluation to (surprise!) E. coli. The same thing happened when E. coli virulence tests were done for the IMP components in the imipenem-resistant enterobacterium Serratia marcescens involved in urinary tract infections and wound infections, W. Zhao et al. (2007).

The upshot is that with the IMP-1 family gang on hand, any imipenem that is forced to trickle through a P. aeruginosa membrane (or counterparts in Serratia marcescens or a suitably mutant E. coli) would run a greater risk of encountering the circulating MBL zinc mines the natural lottery had brought on the scene (possible outcome: you may die from whatever infection the imipenem isn’t helping you with anymore). Some “limit to Darwinian evolution” that turned out to be.