Lighting up Leptospira interrogans to test antibiotic treatment of chronically-infected mice

A group at the Pasteur Institute succeeded in generating a bioluminescent strain of Leptospira interrogans.  Their study was published last year in PLOS Neglected Tropical Diseases.

The benefit of having a bioluminescent strain is that infections of small laboratory rodents can be monitored without sacrificing the animals.  Instead, the animals are placed in a special whole-body imager that detects light emitted from the bodies.  The location of the infection in the animals can be determined from the images, and the light intensity measured by the imager gives an idea of the bacterial load in infected tissues.  After imaging, the animal can be returned to its cage, and additional images of the same animal can be taken later as the infection progresses.

Another advantage of using bioluminescent bacteria is that the luciferase reaction requires ATP, meaning that the bacteria must be metabolically active to light up.  Dead bacteria containing luciferase will not generate light, unlike green fluorescent protein (GFP), another reporter used to image bacteria (see this post that describes an experiment with Borrelia burgdorferi expressing GFP).

To generate bioluminescent L. interrogans, the researchers hooked the firefly luciferase gene up to a strong L. interrogans promoter and inserted the construct into a transposon carried on a suicide plasmid.  The plasmid was then introduced into L. interrogans by conjugation (see this blog post for details of the process).  Cultures of the engineered spirochetes lit up when luciferin, the luciferase substrate, was added.  The amount of light emitted depended on the culture density: more light was detected at higher culture densities.

Mice are ideal models to study persistent infection of the kidney since many rodents are chronic carriers of Leptospira out in nature.  These rodents don't get sick from the infection, but they contaminate the environment with the spirochete every time they urinate.

To see how chronic infection is established, the investigators injected a sublethal dose of 10⁷ bioluminescent L. interrogans cells into the abdominal cavity of C57BL6/J mice and took sequential images of the animals over the following months.  Albino mice were used because dark fur blocks the signal emitted by the bioluminescent bacteria.  (They later showed that standard C57BL6/J mice with black fur could be used as long as they shaved the fur off before placing them in the imager.)  The mice were injected with luciferin 10 minutes prior to imaging.

There turned out to be two phases of infection (see the "MFlum1" plot in the graph below).  In the acute phase, the bioluminescent signal rose to a peak by day 4 and quickly declined to background levels by day 7.  The signal then started increasing again slowly and plateaued after a month.

Figure 2A from Ratet et al., 2014.  Images of a single mouse taken sequentially are shown below the graph.  Click for larger image.  Source.

Images of a single mouse taken at different times after inoculation are shown below the graph.  Thirty minutes after inoculation, signal was detected in the abdominal cavity.  By day 3, the signal consumed the entire mouse.  At this point, the bacteria were probably circulating in the bloodstream.  By day 6, the signal was almost completely gone.  After day 6, the signal appeared again, but it was confined to the kidneys.  The intensity of the signal in the kidneys increased with time.  They did not detect signal anywhere else in the animals during the second phase.  They even sacrificed some of the infected mice 2 months into the infection to check the organs directly, but they failed to detect Leptospira by bioluminescence and qPCR in the brain, lungs, spleen, liver, or blood.  Not surprisingly, bioluminescence was detected in urine, confirming that the mice were shedding live L. interrogans.

Next, the investigators tested the effectiveness of antibiotics in treating mice infected with the bioluminescent L. interrogans.  Several antibiotics are used to treat acute leptospirosis in humans, including penicillin and azithromycin.  It is generally believed that antibiotics are more effective if provided early in acute illness.  Therefore, the investigators tested whether the timing of antibiotic treatment was important for effectiveness.

As expected, penicillin treatment was most effective when treatment was started at the beginning of the acute phase.  In mice treated with daily injections of penicillin for 5 days starting a day after infection, no bioluminescence was detected in the kidneys, and urine was free of L. interrogans as measured by qPCR.  However, if treatment was delayed until three days after inoculation, during the peak of acute infection, a low level of L. interrogans was detected in urine by qPCR even though no bioluminescence was detected in the kidneys.  It is likely L. interrogans was present in the kidneys but at levels too low to be detected by the imager. The bioluminescence approach clearly does not have the sensitivity of qPCR.  Additional experiments revealed that the limit of detection was 100 bioluminescent L. interrogans cells in 100 μl of buffer.

Penicillin was even less effective when administered after the spirochetes settled in the kidneys. When penicillin treatment was initiated at peak bacterial load in the kidneys, day 25 of infection, the signal diminished by over 90% but then bounced back to the level observed before treatment began (see figure below).  Ciprofloxacin also failed to eradicate the bacteria.

Figure 5A from Ratet et al., 2014.  Antibiotics were administered for 5 days started on day 25 of infection.  Cipr, ciprofloxacin; Pen, penicillin.  Source.

On the other hand, azithromycin managed to knock the signal in the kidneys down to background levels (see graph below).  However, the signal came back within a week, although not to the high levels seen in untreated mice.  A second course of antibiotics starting on day 112 knocked the signal back down to near background levels, but again, spirochete numbers rebounded, although not to the levels seen before retreatment.

Figure 5B from Ratet et al., 2014.  Azithromycin was administered for 5 days starting on day 25 and day 112 of infection.  Source.

Why are antibiotics ineffective in eradicating L. interrogans during the chronic phase?  Like other bacteria, L. interrogans can form biofilms in vitro.  Scientists who work with Leptospira believe that they also assemble into biofilms within the kidney tubules during chronic infection.  Biofilms are hard to eliminate in part because they harbor persister cells that tolerate antibiotics.  (See this post for some background on persister cells.)

I should caution readers from concluding that tolerance accounts for the poor effectiveness of antibiotics in treating human cases of acute leptospirosis.  As the authors point out, leptospirosis patients die because the infection severely injure vital organs.  By the time lethal damage occurs, it does not matter whether antibiotics kill all of the spirochetes.

So does the mouse model have any relevance to human leptospirosis?  The authors argue that asymptomatic carriage of Leptospira has been overlooked.  A 2013 study from the Netherlands revealed that 21% of patients who contracted  leptospirosis continued to suffer from headaches, muscle aches, and extreme fatigue two years later.  This may reflect unrepaired tissue damage inflicted during acute infection, but no one checked for the presence of Leptospira in these patients.  Another study from Peru (see this post) describes asymptomatic individuals who may have persistent Leptospira infection. Kidney function was not checked in the Peruvians, but there is reason to believe that chronic infection affects the kidneys despite the lack of symptoms.  Mice chronically infected with L. interrogans are not visibly sick, but they end up with scarred kidneys (fibrosis), as explained in this study.

If persistent asymptomatic infections really do occur in humans, it may be sensible to treat with antibiotics.  The chronically-infected mouse will serve as a nice model for testing antibacterial regimens that target Leptospira living in the kidneys.

References

Ratet G, Veyrier FJ, Fanton d'Andon M, Kammerscheit X, Nicola MA, Picardeau M, Boneca IG, & Werts C (2014). Live imaging of bioluminescent Leptospira interrogans in mice reveals renal colonization as a stealth escape from the blood defenses and antibiotics. PLoS Neglected Tropical Diseases, 8 (12) PMID: 25474719

Goris MG, Kikken V, Straetemans M, Alba S, Goeijenbier M, van Gorp EC, Boer KR, Wagenaar JF, & Hartskeerl RA (2013). Towards the burden of human leptospirosis: duration of acute illness and occurrence of post-leptospirosis symptoms of patients in the Netherlands. PloS One, 8 (10) PMID: 24098528

Fanton d'Andon M, Quellard N, Fernandez B, Ratet G, Lacroix-Lamandé S, Vandewalle A, Boneca IG, Goujon JM, & Werts C (2014). Leptospira Interrogans induces fibrosis in the mouse kidney through Inos-dependent, TLR- and NLR-independent signaling pathways. PLoS Neglected Tropical Diseases, 8 (1) PMID: 24498450


Related posts

• Life after leptospirosis, a pilot study
• Healthy human carriers of the spirochete Leptospira in the Peruvian Amazon
• The magic of antibiotic tolerance

Do Lyme disease spirochetes produce a toxin?

According to the current view of Lyme disease pathogenesis, tissue damage is caused by the inflammatory response to the spirochetes.  Borrelia species do not produce toxins that injure the host directly.  A new study published in BMC Microbiology may force us to modify our view.

The study shows that some Borrelia strains carry a set of genes with the potential to generate a peptide resembling streptolysin S (SLS), a potent toxin produced by the pathogen Streptococcus pyogenes.  The enzymes that produce SLS in S. pyogenes are expressed from a cluster of genes surrounding sagA, a tiny gene encoding the SLS precursor.  The peptide produced from sagA is nontoxic; it has to undergo several alterations to its structure to become toxic.  A critical modification is carried out by the SagBCD protein complex, which converts the side chains of cysteine, serine, and threonine into ring structures.

Figure 2 from Molloy et al., 2011

Other genes surrounding sagA encode a peptidase that is thought to trim the leader peptide from the amino terminus of the SLS precursor and an ABC transporter that may be responsible for expelling SLS from the cytoplasm.

Figure 1A from Molloy et al., 2011

SLS targets neutrophils and possibly other immune cells during S. pyogenes infection.  SLS-like toxins are also produced by other Gram-positive pathogens, including Staphylococcus aureus, Listeria monocytogenes and Clostridium botulinum.

The investigators mined the genomes of other bacteria in search for genes encoding the machinery that generates SLS-like toxins.  They found SLS-like gene clusters in various Firmicutes and Actinobacteria, both Gram-positive groups of bacteria.

The researchers also found the gene cluster in the genomes of Borrelia afzelii strain PKo, Borrelia valaisiana strain VS116, and Borrelia spielmanii strain A14S.  B. afzelii is a major cause of Lyme disease in Europe and Asia.  B. valaisiana and B. spielmanii are responsible for occasional cases of Lyme disease.

Figure 4 from Molloy et al., 2015.  Top: organization of SLS-like gene cluster in S. pyogenes and three Borrelia strains. Bottom: sequence of the SLS precursor (SagA) and the borrelial SLS-like precursors.

They also used PCR to screen the DNA of 140 patient and tick isolates of Lyme Borrelia for the genes encoding the SLS-like biosynthetic machinery.  Most of the isolates were obtained from Europe and the U.S., with a few coming from Asia.  Design of the PCR primers was based on the sequence of the B. valaisiana bvalB, bvalC, and bvalD genes, which encode homologs of the S. pyogenes sagB, sagC, and sagD gene products.  Most of the B. garinii, B. afzelii, B. valaisiana, B. spielmanii, and B. lusitaniae isolates that were examined tested positive.  On the other hand, none of the 22 isolates of B. burgdorferi or 13 isolates of B. bavariensis were PCR positive.  These results indicate that SLS-like sequences are widespread among Lyme disease spirochetes (though not in B. burgdorferi).

The next step was to show that the SLS-like borrelial gene actually encoded a peptide that damages mammalian cells.  A simple assay based on the ability of many toxins to rupture (hemolyze) red blood cell in vitro is available.  Hemolysis is measured easily by mixing the toxin with sheep red blood cells.  Hemoglobin released from the ruptured cells is quantified with a spectrophotometer.

They decided to test the SLS-like peptide encoded by B. valaisiana, BvalA, for hemolytic activity.  The researchers succeeded in expressing and purifying a recombinant form of BvalA.  Not surprisingly,  BvalA was not hemolytic because its amino acid side chains had to be converted into ring structures necessary for the peptide to injure red blood cells.  They wanted to mix BvalA with the BvalBCD protein complex so that the peptide would be modified, but they could not generate the protein complex.  Instead, they used the SagBCD complex from S. pyogenes to modify the BvalA peptide.  When they did this, they finally observed hemolytic activity.

Red blood cells are unlikely to be a major target of borrelial SLS-like peptides during infection.  So what is the real target?  More studies are needed to answer this question, but we should consider the possibility that the toxin has nothing to do with Lyme disease.  Instead, it may help the spirochete to survive during its residence within the tick vector.  A number of nonpathogenic bacteria carry gene clusters distantly related to the ones that produce SLS.  Several peptide toxins produced by these bacteria are known to kill competing microbes.  Like us humans, ticks have a microbiome inhabiting their gut.  Some Lyme spirochetes may need to secrete the toxin to ward off their microbial neighbors.

References

Molloy EM, Casjens SR, Cox CL, Maxson T, Ethridge NA, Margos G, Fingerle V, & Mitchell DA (2015). Identification of the minimal cytolytic unit for streptolysin S and an expansion of the toxin family. BMC Microbiology, 15 PMID: 26204951

Molloy EM, Cotter PD, Hill C, Mitchell DA, & Ross RP (2011). Streptolysin S-like virulence factors: the continuing sagA. Nature Reviews Microbiology, 9 (9), 670-81 PMID: 21822292

A biosignature of early Lyme disease

Laboratory testing for Lyme disease involves two-tier antibody testing with sera from patients suspected of having the disease.  The first step is usually an ELISA with a cell lysate of Borrelia burgdorferi as antigen.  If the ELISA results are positive or borderline, a Western blot is done to confirm that the patient has Lyme disease.  Direct detection of Borrelia burgdorferi by culture would be the preferred laboratory test, but it takes too long for the spirochete to grow.  Culture of patient specimens is done only for research studies.

Source: CDC

In general, the problem with antibody testing for infectious diseases is that it takes time for the immune system to generate antibody against the pathogen.  Therefore, patients in the early stages of infection may test negative.  False-negative tests may delay appropriate treatment until the illness worsens.  For these reasons, scientists have been trying to come up with better laboratory tests for infections whose diagnosis relies on detecting antibody against the infectious agent.

One new approach being developed for a few pathogens involves measuring the amounts of each of the thousands of small molecules found in the sera of infected patients.  This is done by liquid chromatography/mass spectrometry (LC-MS), which accurately and precisely measures the size of small molecules, even in complex substances like serum.  The assumption is that the composition of small molecules (the so-called "metabolome") starts to change in a predictable manner as soon as someone is infected.  The metabolome changes because tissues react to the pathogen by generating inflammatory molecules that leak into the bloodstream.  Another critical assumption is that the changes that occur in the metabolome are unique to each pathogen.  Analysis of the patient's metabolome may therefore allow clinicians to quick diagnose any infectious disease whose metabolome has been characterized.

A recent CDC study revealed the metabolome of early Lyme disease.  The investigators obtained sera from 89 patients who had early-stage Lyme disease.  All had erythema migrans (EM), the rash characteristic of Lyme disease.  Most were also culture or PCR positive for Borrelia burgdorferi.  The patient sera were compared with sera from 50 healthy individuals by LC-MS.

After two runs of LC-MS with each sample, the researchers identified a set of 95 small molecules whose levels consistently differed between patients with early Lyme disease and healthy individuals.  Statistical modeling of the data allowed the investigators to refine the biosignature to a set of 44 molecules that identified Lyme disease in the 139 (89 + 50) subjects with the highest sensitivity and specificity.

To better gauge the performance of the biosignature in identifying those with early Lyme disease, the investigators conducted LC-MS on sera from another group of 91 patients shown to have early Lyme disease by the same criteria as the first set of patients.  Control sera came from 108 healthy individuals.  Another set of control sera was obtained from 101 patients with other diseases that could be confused with Lyme disease clinically, serologically, or microbiologically: syphilis, severe periodontitis, infectious mononucleosis, and fibromyalgia.  All patient and control sera were also tested by the standard two-tier antibody test.

The sensitivity of LC-MS testing turned out to be much higher than that of two-tier testing: 88% vs. 44%.  The specificity of LC-MS was 94% with healthy sera and 95% when sera from patients with other diseases were tested.  These values were not significantly different from the specificities of 100% and 95% achieved with two-tier testing.

These results show the promise of using the metabolic biosignature to help diagnose early Lyme disease.  However, note that all patient sera used to uncover the biosignature and assess its performance came from individuals with EM.  In practice, a clinical diagnosis involving the classic bulls-eye EM with a patient history suggestive of Lyme disease does not require confirmation by laboratory testing.  Patients without EM are more likely to need laboratory testing.  According to the CDC, 20-40% of Lyme disease patients do not have EM.

To get an idea of how well the biosignature performs on patients without EM, the investigators obtained sera drawn from 22 cases with early Lyme disease who tested positive with the C6 ELISA, a newer antibody test.  The antigen for the C6 ELISA is a highly-conserved peptide from the B. burgdorferi surface protein VlsE.  Eight of the 22 patients did not have EM.  The EM status was unknown in another eight patients.  The remaining six patients had EM.  Unfortunately, the results for each subgroup were not presented by the authors, so we don't have a firm answer about the performance of LC-MS testing on patients without EM.  What we can say is that even though more than a third of the 22 patients did not have EM, the sensitivity of LC-MS testing remained high at 86%.  In contrast, the sensitivity of two-tier testing with this group was only 41%, even though the investigators stacked the deck by using the C6 ELISA as the first tier with this group.  Future testing of the biosignature should include a larger number of sera from EM-negative patients in the early stages of Lyme disease.

As discussed in the paper, sera from patients with skin conditions that could be confused with EM (e.g., STARI, cellulitis) should be examined in future studies to make sure that the early Lyme biosignature can be used to rule out those conditions.  The authors recommend that sera from patients with neurologic, cardiac, and arthritic forms of Lyme disease also be examined to see if biosignatures specific for these more serious forms of Lyme disease could be identified.

Reference

Molins CR, Ashton LV, Wormser GP, Hess AM, Delorey MJ, Mahapatra S, Schriefer ME, & Belisle JT (2015). Development of a metabolic biosignature for detection of early Lyme disease. Clinical Infectious Diseases, 60 (12), 1767-1775 PMID: 25761869