‘Subway Map’ Approach To Finding Targeted Therapeutics For Lyme Disease
By Deborah Borfitz
November 7, 2023 | Scientists in Boston are flipping the script on the current antibiotic development paradigm by painstakingly choosing a good metabolic target for combatting Lyme disease. The typical approach has been to test vast numbers of chemicals against the culprit bacteria rather than design a compound to specifically go after one of its essential biological activities, according to Peter Gwynne, a molecular and microbiologist at Tufts University School of Medicine and researcher with the Tufts Lyme Disease Initiative.
As reported recently in mSystems (DOI: 10.1128/msystems.00835-23), Gwynne and his team developed the first-ever genome-scale metabolic model, or computational “subway map,” of key metabolic activities of the Lyme disease bacterium Borrelia burgdorferi. The bacterium has a small genome and limited metabolism, making it an ideal candidate for the development of narrow-spectrum antimicrobials, says Gwynne, a recent recipient of the Emerging Leader Award from the Bay Area Lyme Foundation.
The “rational design” tactic may have fallen out of favor, he says, but it avoids the downside of relying on broad-spectrum antibiotics that contribute to the development of antimicrobial resistance. Many patients, particularly if they have chronic Lyme symptoms or recurring Lyme disease, would also welcome a targeted antibiotic they could take for years without any of the unwelcome side effects.
New business models are needed to make it financially feasible for companies to make antibiotics that are only able to kill or inhibit a limited species of bacteria, Gwynne adds. Many clinicians today don’t know the bacterial infections of their patients at the species level because they don’t have to. “All the treatments are the same.”
B. burgdorferi is a notoriously “weird” bacteria that Gwynne began to study out of scientific fascination, and in earnest once he moved to New England where practically everyone knows someone who has had the tickborne disease, he says. Being across the street from Tufts Medical Center, he also has occasion to work with clinicians who want but have neither Lyme disease vaccines nor good diagnostics.
As a bacteriologist, he also dislikes standard treatment with a broad-spectrum antibiotic that causes all kinds of off-target problems. Outside of the academic field, the idea that bacterial infections should be treated the same as any other disease—based on the diagnosis—seems to have limited acceptance, says Gwynne.
The go-to treatment for many doctors is an antibiotic that acts against an extensive range of disease-causing bacteria, the clinical rationale being that a proper diagnosis is unnecessary if you can just give everyone the same drug. But that disrupts the human microbiome in ways that heighten the risk of harmful, long-term conditions that impact human health, says Gwynne, emphasizing the need to “radically redesign the pipelines” toward narrow-spectrum antibiotics. Current therapeutic practice creates antibiotic-resistant bacteria that are hard to treat, in addition to causing side effects such as diarrhea or rash.
Lyme disease is hard to detect just on a fundamental level, he quickly adds, and a reasonable amount of funding has gone into trying to develop new diagnostics. Like cancer, Lyme disease appears to be many diseases in one based on the variety of presenting symptoms that include fatigue, brain fog, swollen joints, and general malaise.
“It really is a complicated area, and it is also a catch-22 because we are trying to develop a diagnostic test and who do you test because you don’t know who has what disease because there is no diagnostic test,” says Gwynne. “Everyone is fighting the fact that we’re trying to build out against nothing... most bacteria were discovered in the mid- to late 1800s and we have been studying them since then. Borrelia burgdorferi was only isolated in the late 1970s.”
In the latest study, Gwynne and his colleagues used big data and machine learning to build a subway map that comprehensively annotates 208 metabolic reactions to 151 genes and serves to demonstrate the possibility of developing treatments that only block the Lyme disease bacterium. The genome of B. burgdorferi, which was sequenced about two decades ago, was run through a bunch of predictive algorithms to generate a model describing the various routes being taken by the bacteria, he explains.
The destinations are known because they’re shared by all biological organisms—e.g., the generation of DNA, proteins, lipids, and cell membranes, he continues. Investigators then went looking for the different “subway lines” to get to those end points.
“Once you find the important metabolic pathways, you can identify the ones that are absolutely necessary... the two or three steps that are completely indispensable,” says Gwynne. “Where there are no alternatives, you wind up at the checkpoints that everything has to flow through.”
In terms of model building, it helped that the genome of B. burgdorferi is only about one-third the size of many other common bacteria, such as Salmonella enterica and Escherichia coli, he adds. A metabolic map of the human genome, by comparison, takes in 13,543 enzymatic reactions.
A Way Forward
Using their subway map, the Tufts team succeeded in identifying two existing small-molecule inhibitors—an anticancer drug with significant side effects and an asthma medication no longer on the market—which selectively kills B. burgdorferi in culture. While the compounds are impractical to use clinically in treating Lyme disease, the essential processes used to identify them represent a way forward in the development of narrow-spectrum antimicrobials, says Gwynne.
Researchers started out with 77 pivotal metabolic pathways and then narrowed the list to the 28 specific to B. burgdorferi. The two small molecule drugs selected to be published in the paper was a “judgement call,” he says, since they were both relatively well characterized, soluble and stable, and could be purchased for a reasonable price. “We collected data from 16 small molecules and found 11 of them killed the bacteria... [and] the data all kind of looks the same.”
The decision now is how to make a rationale choice about which of the remaining candidates to investigate from the list of 28, says Gwynne. “If I had more hands, I’d take forward all of them.” Ultimately, only one or two will likely survive aggressive laboratory testing.
The odd metabolism of the Lyme bacterium has separately inspired Gwynne and his team to examine whether people with chronic Lyme symptoms are still infected or instead suffering from an immune malfunction, he reports. The metabolic pathways of B. burgdorferi are few because it steals what it needs from the host, including phospholipids, he says.
Phospholipids, presented to the body by the bacteria, might be driving antibody formation in the host, speculates Gwynne. The team is now investigating the scavenging of host cell components behind this phenomenon.
Auto-antibodies are of course implicated in all kinds of diseases, including fibromyalgia and multiple sclerosis, which share a lot of the long-term symptoms of Lyme disease, he says. But as has already been seen, not every Lyme disease patient has these antibodies.
Similar subway maps could be built for multiple types of bacteria with relatively small genomes, such as those that cause syphilis, chlamydia, and Rocky Mountain spotted fever, says Gwynne, adding that his training is in experimental approaches and not computational predictions. But experiments aren’t a practical option with pathogens that grow slowly, are highly host-dependent, require special culture conditions, or aren’t particularly well studied.
While it takes a week to genetically engineer E. coli (to, for example, knock out a gene), the process for B. burgdorferi takes a couple months, he offers as an example. And Chlamydia trachomatis is hard to grow because it lives inside mammalian cells. The idea here is not to replace but streamline the traditional experimental approach at the hypothesis generation stage, says Gwynne.
Many people are using genome-scale metabolic models for drug discovery in the biotech field as well as to produce high-value chemicals such as ethanol, which is the opposite of the subway map approach of Gwynne and his colleagues. Instead of trying to kill bacteria, they are trying to “push the metabolism toward overproduction,” he says.