Killing the Unkillable

Without new and transformative drugs, tuberculosis will persist.

Mycobacterium tuberculosis (TB)

Adapted from Catching Breath: The Making and Unmaking of Tuberculosis by Kathryn Lougheed. Copyright © Kathryn Lougheed 2017. Published by Bloomsbury Publishing Plc. Reprinted with permission.

Tuberculosis wins by waiting. Ninety percent of cases are asymptomatic but these infections are a ticking time bomb, waiting for the opportunity to go off. When they do, it takes a six-month course of antibiotics to cure the infection. The huge worldwide reservoir of latent disease and the painfully long treatment times are one of the big problems in TB control. One of the solutions could be found in the pathogen’s ability to form persister cells that can survive the bad times and bounce back once the threat—be it a person’s immune system or antibiotic challenge—has passed.

Persister cells are not unique to Mycobacterium tuberculosis (TB); they are a global phenomenon employed by all species of bacteria. The formidable penicillin expert, American microbiologist Gladys Hobby, was the first to identify these cells in 1942, but Irish scientist Joseph Warwick Bigger coined the term “persister” in 1944. He showed that penicillin kills only 99 percent of a staphylococcus culture, leaving behind a stubborn 1 percent that somehow survives. Bigger’s cells weren’t genetically different from the 99 percent, but they were what’s termed “phenotypic variants.”

From my undergraduate biochemistry notes, now 10 years on, I still recognize biphasic kill curves. The number of survivors are on the Y-axis and time is on the X-axis. Add the weapon of death, and the curve drops from its starting point at 100 percent survivors and keeps going down. And then something happens. The line starts to level off like the tail end of a ski slope. The rate of death slows, and while bacteria are still dying, they’re taking their time about it. For example, in mice infected with TB, the antibiotic isoniazid causes a relatively rapid decrease in the number of bacteria in the lungs over the first four weeks. And then the killing grinds to a halt. Maybe 1/1000 or 1/100 of the initial population is left, and they take much, much longer to die.

The reason is that not every cell is doing the same thing. The cell population in a laboratory culture or in a mouse lung is not a homogenous mixture of identical individuals. All appear the same, but when it comes to how they live their lives, not all follow the same path. When microbes start to see the warning signs of impending doom, more and more of them change their behavior as a survival strategy. A study done by Galina Mukamolova and her research group at the University of Leicester showed that bacteria adapted to life within the lung are more tolerant to first-line TB drugs than cells grown in the lab. A programmed adaptation to the intracellular environment most likely ensures the survival of at least some of the bacteria, and also protects them from being killed by antibiotics.

WPA Federal Art Project, 1936–1939

Library of Congress
In 2015, a paper from John McKinney and colleagues, at the McKinney Laboratory of Microbiology and Microsystems in Lausanne, Switzerland, looked at what happens to populations of M. tuberculosis when they hit upon hard times. By engineering M. tuberculosis to make green fluorescent protein during periods of active growth—the more growth, the more fluorescence—McKinney’s group was able to visualize different populations of bacteria based on growth rate. In a culture flask, they found that everything looked similar in terms of greenness. But when they exposed the bacteria to nutrient starvation, the diversity started to increase. McKinney’s team decided to test their bacteria under the most extreme conditions—a real-life infection where they would have not only nutrient depletion to contend with, but also an attack from the host’s immune system. When they looked at the bacteria clinging on to life in mouse lungs, they saw a huge range of phenotypic heterogeneity. Some cells were growing actively, others were ticking over slowly. A subpopulation of cells was non-growing but still metabolically active. Alive, but only just. These cells were absent in mice engineered to be lacking in a key component of the immune system, indicating that the formation of these non-growing cells requires immune challenge.

More recently, Samantha Sampson and a group of researchers in the division of Molecular Biology and Human Genetics at Stellenbosch University in South Africa employed a similar technique, only this time they engineered their M. tuberculosis to make a fluorescent protein that is gradually lost from the cells as they divide. The team infected macrophages with this strain and looked at how the population of cells diversified over time. At early timepoints, the cells were all fairly similar; after 48 hours, the scientists started to see a slowgrowing population appear. As the macrophage infection progressed, an increasing number of bacteria also became tolerant to antibiotic challenge. The big question here is whether the slow-growing population is the same group of bacteria that can survive drug treatment. It would certainly make sense. Most antibiotics, after all, work on actively dividing cells. By choosing to go to sleep, the bacteria take away the weakness that the antibiotic could exploit. It follows that one of the reasons why TB is so hard to treat is partly due to non-growing persisters protected from death. Understanding the pathways used by M. tuberculosis to enter this non-growing state could teach us how to kill this troublesome population.

WHO Global TB Report 2016

A 2011 paper from biologist Kim Lewis and his team at the Antimicrobial Discovery Center at Northeastern University in Boston, MA, looked at the genes switched on in M. tuberculosis persisters. This population was selected for by exposure to the antibiotic d-cycloserine. All the susceptible bacteria die, leaving behind the antibiotic persisters. Among the genes upregulated in these persisters were 10 toxin-antitoxin modules. Toxin-antitoxin systems are a yin and yang of the bacterial world. One protein—the toxin— inhibits vital functions within the cell; the other—the antitoxin—neutralizes the toxin. By shifting the balance of the two proteins, a cell can regulate its own growth and shift into a non-growing state, if required. M. tuberculosis, interestingly, contains a whopping 79 toxin-antitoxin pairs, suggesting that it probably uses them for something very important. In a more recent U.S.-Korea collaboration, Lewis took a different approach and isolated mutants of M. tuberculosis with an increased ability to form persisters. The team started with a library containing mutants in every gene that could be mutated, then treated the mixture with antibiotics. If the mutant yielded fewer persisters, the mutated gene was vital for persister formation. If it generated more persisters—a super-persister—then the mutated gene was inhibiting persister formation. The genes identified came in a variety of flavors, including various machinery involved in PDIM biosynthesis, phospholipid biosynthesis, metabolic pathways, RNA maturation, and various other processes. Among the genes switched on in the super-persisters, once again there were several toxin-antitoxin modules. In addition, the scientists found genes involved in lipid biosynthesis, carbon metabolism, and various transcriptional regulators. A bit of everything, basically. All of this suggests that M. tuberculosis persister formation involves multiple pathways containing lots of redundancy—different pathways and proteins that overlap in function (the 79 toxin-antitoxin systems are a case in point). 

In parallel to his work on understanding the mechanisms of tolerance, Kim Lewis is also interested in methods of killing persister cells. But how do you kill something with a dozen or more backup options? “Apparently, we’ve hit upon a very interesting Achilles heel, not only in TB, but in bacteria in general, and that is the Clp protease,” he told me. His drug, lassomycin, forces this protease to break down adenosine triphosphate (ATP), the energy currency of life. No matter whether a cell is growing, hibernating, or something in between, it still needs ATP. Switching everything off can’t protect a cell from a drug that interferes with ATP. Lewis thinks lassomycin— an antibiotic isolated from a soil bacterium—probably evolved specifically to kill persisters. A way for one species of bacteria to carve itself a niche in the overcrowded soil is by killing its neighbors. “This tells us that nature isn’t targeting any mechanism of persister formation, probably because it’s impossible due to redundancy. Lassomycin bypasses the complexity and it simply does something entirely different to hit persisters.” In a 2014 paper from Lewis’s lab, there is one of those biphasic kill curves I mentioned above. Instead of the ski-slope curve, lassomycin results in a straight downward line, suggesting that there is no obvious persister population with tolerance to this drug.

A lethal dose of bactericidal antibiotic added at time zero rapidly eradicates the sensitive bulk of the population (blue) until only nongrowing persister cells (red) that are killed at a slower rate remain. The slower killing has been interpreted to reflect the persister resuscitation rate, but this remains to be substantiated experimentally. The termination of antibiotic treatment enables the population to be replenished by resuscitation of surviving persisters.

Kim Lewis is a strong supporter of the idea that nongrowth is behind the persister phenomenon. Some members of the TB research community don’t agree. For the 60 or so years following Bigger’s work, his idea that persisters were non-replicating or dormant cells persisted. It made sense, so few challenged it. It took until 2004 for someone to finally provide some evidence to directly support the hypothesis. But it wasn’t all-species, all-antibiotics evidence. According to some scientists, switching off and shutting down is not a universal mechanism of persister formation. John McKinney has argued against what he describes as a scientific dogma arising from what was never more than a hypothesis. The big issue with solving this difference in opinion is that persisters are ridiculously hard to work on. These reclusive bet-hedgers are not only extremely rare, but they don’t behave in a way that makes them easy to observe. They’re off the grid, lost in a sea of noise, and avoiding doing the things that everyone else does, waiting patiently instead for the rest of their species to die so they can be proven right.

It wasn’t until recently that science caught up with persisters and came up with ways of spying on them. Microfluidics is sometimes labelled as lab-on-a-chip technology. It involves culturing bacteria in extremely small volumes in a miniaturized device that can be visualized using time-lapse microscopy. Because you’re using tiny, tiny numbers of cells, you can watch what each one gets up to as an individual, rather than basing your conclusions on averages and generalizations. You can start out with, say, 100 cells, then add an antibiotic and watch 99 of those cells die. The one survivor is your persister, trapped in the chip’s single focal plane with all its tricks laid bare. Another persister researcher, immunologist Sarah Fortune, at the Harvard T. H. Chan School of Public Health, says of single-cell imagining that “there’s an appealing concreteness to looking at cells and seeing that they’re different from each other.” Kill curves or microarrays or a myriad of other techniques don’t give you this same window into the secret lives of individual bacteria.

Illuminated tram-car advertising was part of an X-ray campaign against tuberculosis by the city of Glasgow, Scotland, March–April 1957.

Wellcome Library
In one of John McKinney’s time-lapse videos of persisters, a single cell glows green in the middle of a dark field of vision. It’s a Matrix-style set-up, with everything the cell needs to survive being pumped into the chamber and all its waste products being washed out the other side: the ideal conditions for our glowing cell to grow and divide, which is exactly what it does. Slowly, the cell lengthens and pinches off into two daughter cells that stretch away like the creeping roots of a tree, still touching end-to-end until the next division pushes them apart. Then an antibiotic—isoniazid—is added, and the cells stop growing. A little while later, the lights start to go out. One, then another. The green glows vanish to leave behind a dark shadow in the background. This shadow is the empty husk of a dead cell. The other cells keep on dying until all that is left are two determined green blobs. They don’t seem particularly bothered about all the isoniazid; in fact, they continue to divide. From a distance, the rate of new life and death would appear to be balanced. It takes a technique like microfluidics to reveal that not every cell follows the herd. Some can survive. Some can thrive.

The game changing part of these experiments, though, was the observation that these drugtolerant cells were not pre-existing, non-growing cells. They were not dormant, as “dogma” would have us believe, but normal, growing members of the population. Something else was at play. Another video sheds some light on how these cells survive antibiotic challenge. If you look closely, the cells appear to pulse red as a protein called KatG, or catalase peroxidase, is switched on and off. KatG is a mycobacterial protein hijacked by isoniazid to convert the drug into its active form. McKinney hypothesizes that the levels of KatG within a cell could be correlated with its likelihood of being killed by isoniazid. The pulsing came as a surprise, though. It turns out that cells make KatG in short bursts, and it’s during these bursts that it is sensitive to isoniazid killing. Once again, these results suggest that there is no single mechanism for bacteria to become tolerant to antibiotics. Some cells might be of the non-growing, dormant variety; others, the result of stochastic gene expression as seen in McKinney’s videos. Kim Lewis, however, would disagree with McKinney’s assertion that his isoniazid-tolerant cells are persisters in the truest sense of the word. This is more an argument over terminology than science. Lewis uses the term “stochastic resistance;” McKinney hijacks “persister.” Either way, all of it adds up to a lot of reasons why TB can be extremely difficult to treat in the real world.

A 2012 paper from Sarah Fortune’s lab also demonstrates that non-growing persisters are not the only solution to surviving antibiotic treatment. When mycobacteria divide, not all the daughter cells are created equal. Instead, asymmetrical division and ageing of mycobacterial cells can lead to distinct subpopulations of cells with variations in growth rate and antibiotic susceptibility. Fortune’s work showed that the slower elongating progeny are more sensitive than their faster counterparts to rifampicin, another front-line antibiotic. Isoniazid, on the other hand, preferentially kills the faster cells. Her conclusions are that this subtle diversification may contribute to the variable outcomes of TB infection and treatment. She refers to these multiple layers of complexity as “an incredible tapestry of variation in the bacterial population.” Another layer of complexity comes in the form of the antimicrobial efflux pump. Originally, the bacterial version of a bouncer on the door, removing environmental toxins that could otherwise kill a cell, efflux pumps have been repurposed to also remove antibiotics before they can do any damage. A 2011 paper from Lalita Ramakrishnan’s group at the University of Washington suggested that one of the reasons why mycobacteria grown in macrophages become drug tolerant is due to the induction of these efflux pumps. The paper highlights the need to not just focus on non-growing cells when considering persistence, but to take into account the huge role that efflux pumps may play.

A chest X-ray of a patient with miliary tuberculosis, which causes numerous small nodules to form in the lungs.

Department of Energy, Yale Rosen/Flickr (CC BY-SA)

How do we develop drugs to kill the (almost) unkillable? Back in the 1940s, we started with a method that gave us our first TB drug, streptomycin. That method has yielded the vast majority of our current antibiotics. Then came the advent of genomics. Genomics was meant to revolutionize drug discovery. The M. tuberculosis genome was published in 1998, and suddenly we had the entire blueprint for this killer pathogen. All we had to do is read it and understand it, and the bacterium’s secrets would be ours. It did spawn an entire new era of TB research—genetic manipulation, transcriptomics, proteomics, structural genomics, comparative genomics—none of these would have been possible without the DNA sequence of M. tuberculosis. But when it came to drug discovery, it wasn’t the future everyone had hoped for. The idea was that a scientist could start out with a gene that looked like it might make an important protein. Genetic manipulation techniques could confirm that the bacterium can’t live without the target protein. From there, it was just a case of developing a drug capable of inhibiting it. This process involves first making and purifying the protein and finding a way to measure its activity. You can then use a gigantic robot to seed 364-well plates, toss in a huge number of potential inhibitors and see if any of them stops the protein from working. These inhibitors might be natural products or compounds that have been synthesized based on existing compounds, often as a result of other drug-discovery endeavors (for example, the hunt for cancer therapeutics). Once you find one that works, you can fiddle with its structure to hopefully improve its activity and other properties required for it to be a good drug (such as its toxicity against humans, how well it is absorbed into the body, and its stability).

A population of genetically identical bacteria includes individuals that express an antibiotic-activating enzyme (green) and others that do not. This ‘individuality’ of bacterial behavior allows some cells to evade killing by the antibiotic.

Yuichi Wakamoto/Neeraj Dhar/John McKinney
Of course, at some point you need to check that it kills bacteria. At the National Institute for Medical Research, outside of London, I was the sole TB microbiologist among a massive team of chemists and biologists, all of whom had put in an extraordinary amount of work into developing assays and using these assays to find compounds capable of inhibiting our protein of interest. Our compounds were brilliant at inhibiting the protein by itself. When it came to M. tuberculosis, though, it only died a little bit, and not enough for the compounds to be of any use as new drugs. I tried to work out why. Was it the super-thick cell wall stopping the compounds from getting inside? Was the bacterium pumping the compounds back out? Did the compounds work better when the bacterium was contained inside a macrophage? Was the target not that good a target? At some point, the TB drug-discovery world shifted. Many people started saying how cell-based approaches like the “good old days” were the way forward. M. tuberculosis was just too complicated to rely on a compound targeting just one enzyme.

According to Kim Lewis, much of the issue was the result of the collapse of the successful antibiotic drug-discovery platform that yielded many of our current drugs between the 1940s and 1960s through the mining of soil organisms. The problem was that only 1 percent of soil microbes can be cultured on a Petri dish; the majority don’t grow. Lewis told me that once that 1 percent was exhausted, no new discovery platform emerged to replace it. So, while the few drugs to have been discovered since the 1960s have used a cellbased screening approach, this doesn’t mean that this method isn’t just as failureprone as my target-based attempts. The “good old days” are gone, and it’s not as simple as going back to basics. “All the new compounds that you see in all fields including TB are a complete lottery,” he explained. “One person has a lucky strike and you have one compound. And then that either succeeds or fails, and it usually fails due to numbers—the probability of failure is higher. Without a platform, we are doomed.” His view is that we can’t keep up with the need for new antimicrobial drugs using this hit-and-miss strategy where the misses far outnumber the hits and any success is mostly down to luck. What we need is something new and transformative. Otherwise, we will be giving M. tuberculosis the opportunity to evolve into a drugresistant monster that can no longer be treated by “modern” medicine.--KL