Gates Foundation TB Drug Accelerator Program TB Drug Testing Models

DRAFT ONLY 3/10/09

PART ONE: HISTORICAL BACKGROUND

It was suggested by Benjamin Marten in 1790 that TB could be infectious in nature. Seventy years later Villemin in 1865 inoculated rabbits with liquid removed from a human cavity and noticed the development of lesions. The animals did not get ill because this species is highly resistant to Mycobacterium tuberculosis, but more susceptible to M.bovis. The reasons for the enhanced virulence of M. bovis are not yet unclear. The same year Budd wrote to the Lancet suggesting that TB was caused by "germs… cast off by persons". Koch later agreed, stating that he thought TB was spread by "droplets of phlegm….flung into the air [by coughing]" [ref].

In 1882 Koch presented his data isolating the tubercle bacillus; he also proposed "tuberculin" as a cure, but this was quickly discredited. This was not for naught however, because in 1907 von Pirquet used tuberculin to induce skin reactions that are the basis of the modern day PPD test. Koch used multiple animal species in his research into tuberculosis, including mice, rats, marmosets, guinea pigs, chickens and himself. In his guinea pig studies, he injected TB from patients into the skin on the abdomen of these animals. After two weeks he noted swelling of draining lymph nodes, and by 4-6 weeks the infection had widely disseminated. If he took the surviving animals and reinfected them, a rapid necrotic reaction transpired at the injection site. This was called the "Koch phenomenon", and was taken to indicate expression of anamnestic host immunity; today this term is used more loosely in the context of adverse reactions, both cutaneous and systemic. Koch himself reported that TB disease will be a very difficult to study because there is no animal model that reproduces exactly the human disease.

In that era, the idea that clean fresh air and proper nutrition was beneficial drove the Sanitorium movement. In 1859 Brehmer opened a Sanitorium in Gobersdorf in the Alps; this idea caught on, and by the end of that century there were over 300 sanitoria in Germany alone. Soon thereafter Switzerland became an epicenter [stimulating Mann’s novel The Magic Mountain]. In 1875 the first sanitorium was established in the US, in North Carolina, by Joseph Gleitsmann, but it soon closed due to lack of money. A few years later, in 1884, Edward Livingstone Trudeau opened his Sanitorium in Saranac Lake NY. He himself had showed the first signs of TB on his honeymoon in 1871, and his brother had already died of the disease. His health deteriorated, and he moved to Paul Smith’s in the Adirondacks, where his health gradually improved and he lived for another 40 years. Aware of the idea of sanitoria he constructed cottages for patients with TB in nearby Saranac Lake, and a replica of these buildings ["Little Red"] now stands on the site of the Trudeau Institute. [The Institute, founded in 1964, is across town from the site of the original Sanitorium]. After Koch’s momentous discovery, Trudeau obtained an English translation, quickly became a self-taught bacteriologist, and used Koch’s methods to obtain his own TB bacterial cultures. He had access to a small island ["Rabbit Island"] where in 1887 he inoculated rabbits with his cultures, showing that those exposed to adverse environmental conditions did not survive. A firm believer that one day drugs would be developed for TB in the same way that Ehrlich had described for syphilis in 1910, in the year of his death Trudeau wrote that "he saw no reason" that TB could not be treated with chemotherapy as other diseases such as syphilis" [ref].

This however was an era of failure. As early as 1915, Corper studied cyanide and DeWitt studied gold compounds in experimentally infected animals and could measure the tissue distribution in uninfected and infected tissues. {Corper, 1915 #127}{Dewitt, 1918 #129}In Vaudremer 1913 claimed that extracts of cultures of Aspergillus could "digest" TB bacilli, and this was tested in some hospitals in Paris, with no success. Koch himself had observed that gold [salts] killed TB, and this idea was popular in the 1920’s. This was finally properly evaluated by Amberson in 1931 when he tested a gold preparation ["Sanocrystin"] and found no benefit. {Watenabe, 1927 #154} In fact, it was extremely toxic. In 1923 Calmette bemoaned the "great number of attempts to discover…chemical agents…arresting [TB] in the guinea pig and rabbit…"

But this was also an era of great advances in synthetic chemistry, and soon work in this field, complimented by work in the "natural compounds" field, began to change matters. As early as the late 1920’s, Domagk and a fellow chemist Klarer had discovered a group of compounds called the sulphonamides, which they showed in mice killed virulent Streptococcus, and then found to inhibit TB growth in vitro. Domagk then found, in the late 1930’s, that two of these, sulphathiazole and sulphathiodiazole, had effects in guinea pigs or rabbits infected with tuberculosis. {Takeda, 1951 #150} With the help of fellow chemists he then found that certain thiosemicarbazones inhibited the growth of TB on slants [extraordinarily, much of this work done while the Allies were bombing his town in Germany; he noted "of 65 employees in my laboratory…many are dead…"] As the war ended, Domagk narrowed down his thiosemicarbazone compounds to three, the best of which they called Tibione [TB-one], then changed the name to "Contebin" [thiacetazone]. At that point they considered clinical trials, but they had no idea about dose, toxicity, or even whether the drug could be taken by mouth. A woman at the Hornhide Sanitorium was the first to volunteer; she had lupus vulgaris TB, and showed dramatic improvement. Domagk then attempted to spread word about this drug across Germany, travelling by horse and carriage because of the total lack of petrol [it was another year before train service in Germany began to return to normal]. However, as it became more widely used, side effects, including depression of bone marrow blood cell production, became evident. In December 1947 Domagk received the Nobel Prize for his work on antibiotics [ref].

In 1943 Lehmann described para-amino salicylic acid, stimulated by earlier observations in the early 1940’s that aspirin stimulated oxygen uptake by the TB bacillus. A close friend, Bernheim, had looked at benzoates, finding them to be bacteriostatic, and Lehmann used this information to look for chemicals with similar structures to that of aspirin in what could be regarded as the first applications of the competitive inhibition approach to designing drugs. By the end of the decade PAS had become an integral part of TB chemotherapy [ref].

The discovery of PAS occurred at much the same time as the discovery of streptomycin. Experiments by Rhines in 1932 had suggested that soil contained microorganisms that produced some sort of natural product that inhibited the growth of TB. This was noted by Selman Waksman, a soil microbiologist at Rutgers University. A close colleague and student of Waksman, Albert Schatz, focused his work under Waksman’s direction on a major candidate, the Actinomyces. In 1943 he isolated a colony he renamed Streptomyces, which produced a molecule that became to be called streptomycin [ref].

Born William Gunn in Glasgow, William Feldman [he took his stepfather’s name] moved to Colorado with his parents. At that time 60% of the entire Colorado population lived there because they had TB or had a family member who did [there were large numbers of sanitoria]. His mother cared for such people, and this induced Feldman to devote his life to TB research. His initial interest however was in animal diseases, and he graduated from Colorado Agricultural College, now Colorado State University, in 1927 with a DVM degree. He then moved to the Mayo Clinic, where he quickly met Hinshaw, a TB physician, and together they read of a study at Johns Hopkins by Rich and Follis. In this study, in 1937, guinea pigs had been injected subcutaneously with human TB and treated with sulphanilimide, based on Domagk’s earlier studies with the similar drug Prontosil, with considerable reduction in disease progression in these animals. In 1940 Feldman and Hinshaw found that a derivative, sulphapyridine, had similar positive effects in guinea pigs infected with TB, and soon after showed even better results with a new derivative, Promin. In 1942 they reported data from 36 TB patients in which at least 8 showed marked improvement, the very first time a drug had had a verifiable positive effect in people. Unfortunately, side effects, such as anemia, diminished interest in their findings. Feldman began to plan further clinical trials in Minnesota, but then came a historical moment; he met Waksman. Another contribution of Feldman was a careful description of the methods used. With the feeling at the time that the chemotherapeutic methods were directed at the infected tissues more than at the bacilli themselves. In vivo methods were considered more reliable than in vitro methods. They spent a great deal of effort at describing the care of the animals, the infective inoculum and the preparation and administration of drugs as well as the recording of the results. {Feldman, 1945 #132}

Waksman did not want to use TB in his lab because of the danger, but Feldman had the facilities for this, as well as the virulent isolate H37Rv. Waksman then published his own studies on streptomycin, with little or no mention of TB in his paper. But in 1944 his laboratory sent 10g of the drug to Feldman, who used it to treat four guinea pigs for 55 days after they had been infected with H37Rv. The drug worked spectacularly. The big problem was the limited amount of streptomycin Schatz was capable of producing, so Waksman approached the Merck Company for help. An agreement was reached, and Merck put their top chemists on the project. In early 1945 Feldman and Hinshaw repeated their studies with larger numbers of animals, again with excellent results; [an important outcome of this was their construction of a scoring system to document lesions in guinea pig organs, the first to do so systematically]. They wanted to publish this, but Waksman wanted first to establish in vitro data. This worked as well, so Waksman published this data first, then quickly followed by Feldman and Hinshaw’s animal data [ref].

Feldman and Hinshaw then began to cautiously treat patients with TB, with the first being a young woman given the drug in November 1944, but again the issue of what dose to use was a difficulty, and the drug was given by repeated intramuscular injections which were painful. Nevertheless, they quickly saw positive results, and in 1945 they published results from 34 patients. By 1952, ten thousand reports on streptomycin and TB had appeared in the medical literature. However, over time, as more patients were treated various serious problems arose…bacterial resistance, high relapse rate, allergy to the drug [which killed the author George Orwell], deafness and vertigo [ref].

The discovery of streptomycin soon thereafter fell into acrimony, with Schatz and Waksman, who had jointly filed the patent on the drug, going different ways. Schatz questioned Waksman’s receipt of royalities [he had never himself received any] resulting in Waksman thereafter down-playing Schatz’s [pivotal] contributions. Schatz responded by suing, with Rutgers University offering three per cent royalties as a settlement. In 1952, Waksman was awarded the Nobel Prize for Medicine. In an act of total travesty, Schatz, Feldman, Hinshaw, and Lehmann were all excluded [ref].

In 1948 Feldman developed tuberculosis himself, probably caused by exposure to the H37Rv he had been working with. In an attempt to help him, Hinshaw reasoned that streptomycin by itself was not enough, and his attention was drawn to the PAS developed by Lehmann. {Lehmann, 1946 #140} Thus began the use of combination therapy to treat TB. Only a year later the BMC conducted a trial in the UK that showed that adding PAS considerably reduced the risk of developing streptomycin resistance. Within a decade streptomycin/PAS/INH was commonly used. In studies conducted in England between 1947 and 1958 it was shown that mortality from TB dropped from about 70 per 100,000 to approximately 12 per 100,000, illustrating the impact of chemotherapy within the context of a highly organized medical infrastructure [ref].

One of Waksman’s many students was Rene Dubos, who would subsequently make a name for himself at Rockefeller University for his discovery of the antibiotic gramicidin [ref]. Initially a student of soil microbiology, he met Oswald Avery, who had described the capsule of pneumococci and its role in pathogenesis, and through brilliant work found an extract [tyrothricin] that digested these capsules. His later work concentrated on tuberculosis [his wife Marie contracted the disease and died of it in 1942] and his great contribution was to develop selective media in which TB and BCG could readily be cultured. In addition of course he is revered for his immense ethical and humanitarian contributions to medicine and disease. He died of cancer in 1982 [ref].

In 1944 a Frenchman, Chorine, had studied leprosy in rats, and had injected them with nicotinamide [derived from Vitamin B]. Based on his positive results he tried the same thing in guinea pigs infected with tuberculosis, and noted a reduction in dissemination of the disease after subcutaneous injection. This study was confirmed in 1948 by Kushner and McKenzie at Lederle but was not advanced further [these same scientists then discovered pyrazinamide]. In a visit to New York in 1951 Domagk included these "hydrazones" on a list of compounds he presented at a conference on Chemistry, in the presence of the US chemist Herbert Fox, who himself had made several of these and found them active against TB. In fact, Fox had already begun to test the most promising, a compound called isoniazid, in patients at the Sea View Hospital in New York. This compound had been provided by Roche, but the drug had also been made independently by Bayer and by Squibb, all from the original thiosemicarbazone template, and each company had made isoniazid unaware of the others. However any possibility of a "patent war" quickly disappeared when it was found that two chemists in Prague had described the chemical structure of isoniazid as early as 1912. Because of the earlier work with PAS, no animal model studies were initially done for INH [although Rist in France subsequently performed some in mice], which went straight to human trials with great success [ref]. Pyrazinamide is a derivative of nicotinamide. A prodrug activated to its active metabolite in the host, pyrazinoic acid had no activity at conventional pH and conditions in vitro (apparently also true of its parent compound, nicatinamide). Yet there is considerable activity in laboratory animals and man importantly highlighting the need to include animals early on in the drug discovery phase [ref].

Several other useful compounds followed. Ethionamide was found in 1956, and first shown to inhibit TB in 1959. Two years later ethambutol was discovered in a synthetic chemistry program. Cycloserine is also produced by Streptomyces, and was described in 1955, as was kanamycin, isolated in Japan in 1957. Perhaps the most important however is rifamycin B obtained from Streptomyces. This molecule is inactive, but was synthetically modified in 1957 to produce the drug rifampin.

Currently drug therapy for tuberculosis, of the drug sensitive variety, is highly effective. Typically an induction phase is employed with four first line drugs including isoniazid, rifampin, pyrazinamide and ethambutol. After intensive therapy for 2 months, the regimen is simplified to isoniazid and rifampin to complete a total of six months. The use of directly observed therapy ensures that response rates will be high. The risk of relapse after successful completing therapy is generally less than 5% [ref].

PART TWO: HISTORICAL USE OF ANIMAL MODELS

Koch’s discovery of the tubercle bacillus raised the potential that it could in some way be treated. Given the desperation of the situation at that time [which of course has only improved to a certain degree even today] many drugs and other compounds were tried, in multiple animal models as well as directly in humans. This included a vast array of materials ranging from heavy metals, cyanide, radioactive substances, creosote, dyes to cod liver oil. Not surprisingly, many of these materials were highly toxic [Calmette, 1923].

In fact, animal models are not a modern day invention, far from it as Koch himself used mice, rabbits, guinea pigs and others. For many years, guinea pigs were used as a diagnostic tool for human and experimental animal tuberculosis. Other species used included guinea pigs {Bartmann, 1955 #115} rabbits {Fujihara, 1955 #134}. Steenken and collegues in the 1950s used radiography on rabbits infected with the bovine tuberculosis and could follow them for relapse. Relapse was often with a drug resistant strain and isolates recovered with INH resistance, rats, voles, chick embryos etc. were less pathogenic for guinea pigs. Steenken also demonstrated that there were numerous acid fast positive by culture negative and guinea pig inoculation negative caseous material. The monkey was used in studies of ethambutol {Schmidt, 1966 #101}. The guinea pig showed less activity than the mouse with PZA. {Mc, 1954 #270} {Mc, 1954 #271} {Muschenheim, 1954 #272; Steenken, 1954 #273; Tompsett, 1954 #274} {Malone, 1952 #268}. Animals were important in the study of intermittent chemotherapy. {Bartmann, 1956 #118} {Bartmann, 1957 #119} mostly these were mice that were used given high numbers of animals needed for such a large number of experimental groups.

Ours is not the first group to synthesize a set of laboratory standards for procedures for testing antituberculosis substances in experimentally infected animals [Feldman and Hinshaw, 1945]. They put forward key elements to standardization of the methodology (see below under streptomycin). A substance should be well tolerated by the animal, should be able to reduce the progression of a well established infection and should be active in the spleen, lungs, liver making also likely to work for a latent infection. He highlighted the importance of the number of organisms injected, the virulence and the route of infection. It was felt to be important that treatment begin after infection was very well established. If an insufficient number of bacilli are administered or an attenuated strain or a different route of infection then the results may be inconclusive because many of the controls fail to develop a lethal form of disease. However, if inoculum too big, then the pathologic changes are accelerated and the results too dissimilar from humans. A simple visual scoring system was developed and recorded at each necropsy and a histopathologic scoring system was reported for organs such as lungs, livers and spleens [Karlson and Feldman, 1949].

Two major events occurred in the 20th century. The first was the development of culture methods that allowed direct preparation and enumeration of mycobacterial suspensions. This allowed the bacilli to be properly grown and their biochemical properties defined, and allowed animals to be infected with measured doses. The second advance was in the use of the animal models, with the gradual development of reproducible and validated models that could be used in multiple laboratories. These were not identical, but we can at least regard them as "semi-standardized" {Ogawa, 1951 #141} {Veltman, 1955 #152}. These animal studies became the basis of human chemotherapy {Grumbach, 1965 #6}.

In the first half of the 1900’s great strides were made in culture methods. Early infection consisted of subcutaneous, intramuscular and other routes of inoculation of animals. Trudeau and others used simple aerosol devices: Figure 2 shows the nose and whiskers of an animal in an early aerosol device from the Trudeau Institute.

Also in the early years of the 1900s, Middlebrook invented the 7H line of media became developed. This great bacteriologist Gardner Middlebrook, as well as the eminent microbiologist Rene Dubos. Although Middlebrook is best remembered for 7H9 and 7H10 media, he also performed several classical guinea pig studies. In addition, he made an enormous contribution to the animal model field by designing one of the first aerosol infection apparatus, still used today. Selective and semi-selective media appeared, such as egg and potato based media, Lowenstein-Jensen medium first appeared in 1932. More "minimal" media were developed, such as Modified Sauton’s and Proskauer and Beck medium.

In 1946 a collection of mycobacterial strains was established at the Trudeau sanitorium in Saranac Lake NY, under the guidance of Dr William Steenken. It initially contained BCG and TB strains, but in 1966 it was recommended that the collection contain multiple strain types, and this was implemented in 1968. In 1980, the entire collection was transferred to the National Jewish Center in Denver Colorado. For each isolate in the collection, the species name and strain was recorded; its history and source; date of deposition at Trudeau; its virulence patterns in mice, guinea pigs, and rabbits; and a series of in vitro tests. The latter included colonial morphology, whether the isolate could grow on plates containing hydroxylamine, isoniazid, sodium chloride, and TCH; pigmentation; speed of growth; and favored temperature. Biochemical tests included for amidase, arylsulfatase, catalase, iron uptake, niacin, nitrate reduction, and Tween hydrolysis. Drug susceptibility testing was performed by determining if cultures grew on quadrants of agar impregnated with drugs, rather than the MIC method.

It is obvious however that during this period of time there was absolutely no consensus on how animal models should be used to determine bacterial virulence; everyone had their own methods and stuck to them. In the context of bacteria submitted to the Trudeau collection, virulence testing was performed in Hartley guinea pigs [400g], ICR outbred mice [25g], and New Zealand White rabbits [~2kg]. For TB, pellicle grown bacilli were homogenized in 7H9 medium containing 0.1% Tween-80, sonicated for 10sec, then frozen at -70oC. After plating to determine CFU, guinea pigs were injected subcutaneously with 10⁶ bacilli in 0.2ml; rabbits were infected with 10⁶ bacilli intravenously; and mice given a range of intravenous doses from 10⁵ to 10⁸. In each evaluation three guinea pigs, two rabbits, and ten mice were used.

Each guinea pig was sacrificed at approximately 80 days. Lungs, spleen, and lymph node "involvement" was scored from 0-4 based on "gross appearance" of organs, not histopathology. Rabbits were harvested at 70 days, and the same assessment performed. Scores were then added together; 4 or below equaled "low virulence", 5-8 moderate, and 9-12 high virulence of the organism. Mouse survival was followed over 90 days, this data was then used to calculate LD50s. In addition, the group of mice given the lowest dose at which 50% of the mice lived, were then harvested to use to count CFU in the lungs. These methods, while crude by today’s methods, nevertheless gave a general indication of strain virulence.

It was the discovery of streptomycin and its subsequent successful testing in guinea pigs that probably best represents a major landmark in the animal modeling field, as described above. In fact, even at this early time during the development of this field, Feldman had tried to establish some basic guidelines for the evaluation of a chemotherapeutic agent. He noted several important elements, as follows:

1. The virulence of the strain of TB.
2. Apply the drug at an established point in the infection (6 weeks in the guinea pig model).
3. Confirm the uptake of TB in every animal (by PPD and tissue histology at 6 weeks).
4. Note the prolonged survival of the animal and evidence of histologic improvement with therapy.
5. Note the histological regression of disease and continue to observe the animal off therapy.
6. Check for emergence of drug resistance.
7. Confirm results by others by duplicating every detail of the original work.

In a sense, this list is humbling. Currently, our laboratory is trying to implement these types of processes, but it obvious from the above list that we are re-inventing most of them. As for point #7, Feldman realized the importance of laboratories confirming the results of others, and this is not necessarily being done currently.

The 40’s, 50’s and 60’s saw a large number of studies using mice and guinea pigs to test new emerging drugs. {Nitti, 1967 #91} {Kleeberg, 1967 #139} {Bartmann, 1959 #125} {Veltman, 1955 #151} {Gray, 1967 #94} Including studies into more sophisticated pharmacokinetics. {Kradolfer, 1968 #89} While early studies focused on whether drugs could prevent dissemination and induction of macroscopic -lesions, this gradually was replaced with evaluation of whether the drug could reduce the bacterial load. A major contribution here was made by Mitchison and his colleagues, who demonstrated in the 1970’s that results of drug studies in outbred ICR mice with protocols based on HRZ were very consistent in predicting the human clinical trials results of the British MRC. {Dickinson, 1968 #88} The lack of contribution of ethambutol to the sterilizing activity in guinea pigs {Dickinson, 1976 #18}. Animals have demonstrated differential effect in lungs vs. spleens and may be drug-class specific. {Kanai, 1980 #327}. Guinea pigs of course provided the first breakthrough in drug therapy, but of course these animals were used much earlier in the history of the field, such as inoculation of animals with human sputum as a diagnostic procedure. Mitchison, as an example, developed the Root Mean Virulence assay for guinea pigs, which related lung pathology to animal survival times.

Rabbits have been used for testing drugs, but inoculation methods including injection into the eye which would not likely be permitted today due to animal welfare concerns. A central issue with the rabbit model is whether it can truly be infected with M.tuberculosis. This animal species is very susceptible to virulent strains of M.bovis, and for this reason is usually infected with the Ravenel strain originally isolated in Wisconsin in 1910. In this infection the lungs of the rabbit rapidly caseate and liquefy, killing the animal in 5-6 weeks. This clearly does not happen with the TB strains H37Rv or Erdman, and yet certain laboratories still regard the rabbit as a valid model of tuberculosis. In addition there are other problems; it is expensive, and there are animal husbandry issues including bacterial shedding through the urine. Moreover, rabbits can have intercurrent infections [Pasteurella, Bordetella bronchiseptica (the latter also seen in GP)].

Papers in the 1920’s looked at "albino rats" and found that much higher doses of TB were needed to kill rats compared to guinea pigs. As a result, rats were considered highly resistant, and by implication, not suitable for research. This began to be questioned in the 1960’s, with studies on the emerging field of Immunology noting the relative ease of access to the rat thoracic duct, and hence a major source of lymphocytes. Transfer of these cells into recipients protected them from intravenous challenge in a similar way to that demonstrated in the mouse model, but it was the latter model which subsequently dominated in studies of the T cell response. Although the literature for this model is relatively thin, there appear to be differences in the efficacy of certain drugs [streptomycin, ethionamide, INH] between rats and mice. Although there are currently multiple problems with the use of non-human primates [low availability, cost, concern for ethical issues], it has been used extensively in the past [ref].

In the 1970’s the tuberculosis field using the guinea pig model was dominated by Smith’s group at Madison. Their primary interest was vaccines, and most of their data was based upon the measurement of CFU counts to demonstrate levels of protection. In contrast, any definition of changes in lung pathology or effects of drugs on the model were rarely reported.

But there is a single exception to the rule [Smith DW, Balasubramanian V and Wiegeshaus E. A guinea pig model of experimental airborne tuberculosis for evaluation of the response to chemotherapy: the effect on bacilli in the initial phase of treatment. Tubercle 1991;72:223-231]. In one of his last papers on the topic Smith treated guinea pigs with conventional INH/RIF/PZA and cut out what he thought were primary and secondary [primary lesion-free] lesions from the lungs before plating them. The drugs gave rise to a rapid clearance in the secondary lesions, approaching sterility, but in the primary lesions after an initial linear drop the bacterial load curved away sideways leaving a residual population.

PART THREE: SHOULD WE STANDARDIZE?

Current Day

Mitchison defined the sterilizing activity of a drug as its ability to kill bacilli that persist for long periods during the last months of chemotherapy. He further stated that "a good sterilizing drug reduces the relapse rate after chemotherapy has stopped". These had their roots in the persisters in Mc Cune’s mice. {McCune, 1966 #290} {McCune, 1966 #291} Measured… "Ultimate measure" is the ability of a drug added to a regimen to reduce the relapse rate. A secondary measure is "the proportion of sputum cultures that are negative at 2 months". The inherent problem here is that if there is any relapse at all, the drug regimen is not sterilizing. Adding R to SH does not significantly improve the initial kill curve in mice, but is needed later. In contrast, Z is needed early but not later.

There are multiple issues regarding assessment of new compounds, and not all of these are adequately addressed. PK/PD issues are paramount; does the animal model protocol establish an adequate concentration of the test drug in the blood, what is the rate of metabolism, does the drug bind to proteins in the blood, to what extent does it penetrate the TB lesions [no way to measure as yet], how does the local environment [pH, etc] influence activity, and so on.

A concept that is relatively new is the distinction between drug resistant bacilli versus drug refractory bacilli. Are drug resistant bacilli already present prior to inoculation or are a small number already there? Do a small number acquire this phenotype once exposed to host immunity. On top of this, as the McCune Cornell studies hinted many years ago, bacilli that survive initially chemotherapy in the mouse and now guinea pig models are not resistant but refractory, meaning that if removed and grown in nutrient media, they are now killed by drugs. {McCune, 1956 #276; McCune, 1956 #277}

In the context of the guinea pig model, recent work at CSU indicates that refractory bacteria can be found in the central caseum and in a rim of fibrosis or non-mineralized necrosis [as yet we have not determined what this rim exactly is] just adjacent. Bacteria occur singly, but most are in clumps or clusters. There is considerable accumulation of iron [ferric ions] around these clusters, suggesting the bacteria are secreting siderophores. Most importantly, there is increasing evidence that these clusters may represent a type of biofilm, and Hatfull’s group [Pittsburgh] has found these biofilms consist of bacteria surrounded by a layer of free mycolic acids. It has been proposed that the drug refractory nature of these persisting bacilli is a drug penetration problem, in that drugs are not getting through this extracellular lipid matrix.

Specific Issues

Historically, inbred strains of mice are relatively new. They were initially bred for genetic studies but this usage rapidly broadened leading to discoveries of immense importance, including the identification of MHC molecules by Gorer and then their further definition by Snell fifty years ago. There are now a large number of well-defined inbred strains, and some of these differ considerably in terms of their susceptibility to infection with tuberculosis. At one time this was thought to be controlled by a single locus [Lsh, which became Bcg, which became Nramp] but this is now known to be far too simplistic an explanation [ref].

Should drugs be tested in outbred rather than inbred mice? Outbred mice are a little bit cheaper, but inbred mice have a built-in lack of natural variance because they are genetically identical. The argument has been made that outbred mice such as the widely-used Swiss variety should be used "because humans are outbred", but the fact is that data obtained in inbred mice is no different.

Another issue, rarely discussed, is the age of the mouse. It should be noted that mice are used in immunology studies when "6-8 weeks of age", but one could argue that at this age mice are still not fully immunocompetent. A related issue is body weight and blood volume; how much volume of injectate is too much for a small mouse?

The two most discussed issues however are the route of infection, and the dose. Tuberculosis infection in fact can be given in several ways, but from the point of view of drug testing, a route has to be chosen which is reproducible and changes in the bacterial load can be validated statistically. Some laboratories use the "realistic" low dose aerosol model, but in a sense "realism" is not actually a driving necessity if one merely wishes to see if a drug is active or not.

But obviously the two main models, intravenous or aerosol, do differ significantly. The initial deposition is clearly different. After aerosol the inoculum arrives in the bronchial tree and alveoli. There is a delay before any bacilli get any further; first to the draining mediastinal lymph node cluster via lymphatic drainage and/or carriage by dendritic and other macrophages, then into the blood. The classic literature then describes a "hematogenous spread" to other areas of the lung and to other organs [spleen, liver], but [in the guinea pig model at least] there may be a route of deposition in separate areas of the lung is via lymphatics not blood vessels. Secondly, we now think that a significant secondary route of dissemination is via the stomach, resulting in infection of the mesenteric lymph nodes.

After intravenous infection, assuming the bacterial inoculum is correctly prepared, there is the characteristic 90% liver, 5-10% spleen, 1-5% lung deposition pattern. In terms of the spleen, the bacilli are directly delivered into a lymphoid organ, resulting in rapid induction of both innate and acquired immunity. Some bacilli are deposited in the lungs, but exactly why this happens has never been questioned. In fact it could be completely accidental, with bacilli attaching to and eroding into lung parenchymal tissue.

The usual doses used in mice are ~50-100 bacilli by low dose aerosol and 3-4 log10 for high dose aerosol, but a range of doses have been used in the intravenous model [10⁵ to above 10⁷]. Aerosol doses that are even lower are possible, but are constrained by the inherent variation; it would be nice to deliver just one bacillus into each mouse lung, but the technology to do so just does not exist [this is probably the major reason why we cannot develop a model of "latent tuberculosis" in the mouse].

Higher doses by the intravenous route powerfully and rapidly induce host immunity. Exactly how this impacts drug effects is not known and has essentially been ignored to date in the field. It may be an irrelevance [and we are currently testing these methods side by side] but we would note that the lymphocytic granuloma is far more advanced in the intravenous model versus the aerosol model at the time drug therapy in the mouse normally starts. Moreover the highly lymphocytic granuloma seen in the mouse i.v. model bears little resemblance to the necrotic granuloma seen at the same time in the guinea pig aerosol model.

High doses of bacilli given intravenously rapidly switch on host mechanisms. NK cell subsets, capable of secreting IFN gamma, are triggered, as are gamma-delta T cells, TH17 cells, and even conceivably regulatory T cells responding to the rapid inflammation induced by the high dose. There is a rapid cytokine response as well, including TNF, IL-10, TGF-beta, and other pro-inflammatory molecules. None of these probably influence whether a drug might work, but can we be completely sure?

One further aspect of this issue regards whether a high dose inoculum is needed so that the drug treatment then "selects" resistant clones, which some believe is the basis of the needed long therapy period in people. We would note however that our data shows that very small aerosol doses given to guinea pigs gives exactly the same outcome, suggesting resistance is acquired or, as we now hypothesize, a result of their specific location and local environment in primary lesions.

The second primary issue regards the virulence of the bacilli in the inoculum. This clearly differs from laboratory to laboratory, simply because each grows their own stocks rather than use a single reference preparation. Issues regarding cultivation methods have been the topic of papers for close to a century, but in general similar media are usually used, and pellicle growth is usually used to generate relatively virulent stocks. How often such stocks are replaced also varies; if one makes large stocks [hundreds of tubes] and keeps them frozen then gradual loss of virulence can sometimes be noticed. There is no agreed standard for replacing stocks in the field. No one, to our knowledge, has compared in a head to head fashion a pellicle grown stock to one which is mouse passaged (pending studies in this Gates effort).

When high doses of bacilli are given then there is the risk that some of the drug-treated mice will still die. The older literature contains multiple examples of this. These include studies in which drugs were tested for their ability to improve percentage survival at a certain time after infection [such as 4 weeks], with this data used to calculate a drug concentration giving the best outcome. Often lung "scoring" was used in place of any attempt to determine lung CFU.

In some cases it was impractical to monitor CFU, so other methods were introduced. For example, lung scoring and pathology was combined with survival to a given date with or without drug therapy in guinea pigs to generate the "root mean virulence" assay. There were silocotic rabbits used and scoring systems to show divergent results of chemotherapy. {Ionescu, 1975 #25}

Relapse

The modern era started to address the issue of relapse with the important studies by McCune and others at Cornell University in the 1950’s. Here, these workers tried to address potential latency after drug therapy of mice, but the bigger picture here relates to the clinical success rate of various drug trials. At face value, no current drug regimens are universally "sterilizing" because each trial has reported some degree of relapse in a percentage of patients. Although the issue of reinfection was not always carefully ruled out.

This is also a weakness of the animal studies. Not every study that is published even includes a relapse evaluation, and in those that do, relapse is usually measured in terms of spontaneous rather than induced relapse. This means that a certain number of mice [numbers which statisticians would argue are too low] are left aside for 3-6 months after cessation of treatment to see which show evidence of regrowth [usually involving plating the lung and looking for colonies]. This has two problems. The first is that tuberculosis even in the mouse is a disseminating disease and bacilli may survive in extrapulmonary sites. The second issue is whether the host immune response somehow contributes, possibly by keeping a few remaining bacilli in an unculturable form.

A way to show the latter is to immunosuppress the host, a tactic rarely used. If this is done, and no regrowth is observed, then only then is it reasonable to conclude that the drug regimen was "sterilizing". McCune realized this factor in his pivotal studies half a century ago [ref].

Current information from the mouse model:

The mouse has been really very helpful in informing human clinical trials. The disadvantage is the lack of pathology. There has been much information forthcoming on the hypoxic nature of lesions in the rabbits, guinea pigs and non-human primates. Addressing this local microenvironment is essential to the shortening of TB drug regimens. This is true at least for established infections.

Issues pertaining to MDR/XDR, lack of drug optimization

With the emergence of MDR and XDR TB the need is urgent for new drugs. A major issue regarding the treatment of tuberculosis (TB) is the increasing incidence of multi-drug resistant (MDR) and extensively drug-resistant (XDR) strains, resulting primarily from improper or incomplete treatment regimens. Treatment of active TB using front line drugs lasts 6-12 months, and treatment of MDR-TB can last two years or more. Tolerability of such regimens is difficult and the long duration of therapy contributes to lack of compliance. Lack of adherence to treatment regimens promotes resistance. While new drugs are in the pipeline, it may be many years before they are widely used in TB treatment. One current challenge regarding the treatment of drug-resistant TB is to find optimized regimens using existing drugs that can be studied right now and that have reasonable activity against MDR TB (e.g. moxifloxacin and linezolid); the basic questions being – for a given combination of drugs, what is the optimal dosage for each individual patient to maximize effect and minimize toxicity? What are the differences based on patient characteristics such as age? These issues require a detailed understanding of the pharmacokinetic and pharmocodynamic issues. It is a time of unprecedented need and a time of unprecedented opportunity. Now that the pipeline has begun showing signs of new agents over the last decade, properly predictive models are essential. We will never have a single drug for tuberculosis treatment and in order to get to relevant human treatment we need to look at regimens. This work will focus specifically on agents that will be playing a role in the treatment of MDR and XDR TB.

Strains used: ATCC strain 27294 is the most commonly referenced H37Rv strain and is also known as TMC 102 was derived from existing strain (Derived from E.R. Baldwin's 1905 human-lung isolate H37 by W. Steenken in 1934). They appear to be the same derivation but it is unclear exactly why there are two deposited in ATCC. CSU type strain for our NIH contract is the TMC102 designation. A second strain of H37Rv was reported (ATCC 25618) and was derived from existing strain (derived from E.R. Baldwin's human-lung isolate H37 by W. Steenken) 1934 NY, United States.

Not much is known about different virulence or different drug responses based on specific TB strain used. Trudeau used very crude animal experiments back in the early part of the 19th century so modern look at the type strains in animals can guide us.

Unanswered questions/challenges

The use of animals in research into tuberculosis changed dramatically in the past fifty years. Whereas [in the US at least] animals now are usually kept in mostly pristine conditions, with sterile bedding, nutritious chow, clean water, sentinel animals to check for secondary infections, and so forth, it is probably true to say that such conditions did not always exist in earlier years.

This has of course evolved due to concerns about "animal rights", i.e. minimally acceptable conditions, proper veterinary care, and adequate prevention of pain. Under NIH [OLAW] rules any Institution working with animal models must have an Animal Care and Usage Committee in place. ACUCs perform internal reviews of all proposed animal experiments, as well as conduct facility inspections twice a year. These are not uniform however; each committee can decide independently on approvals as long as these are recorded. As specific examples, most if not all ACUC would not these days permit retro-orbital bleeding of mice. ACUCs also vary considerably in their attitude to studies that monitor animal survival. Some allow it, some require very substantial justification, and some frown on it.

Applications to ACUC committees usually require a statistical power calculation. In this, predicted data is used to establish the statistical "power" of a particular experiment. This tells the ACUC if the study generates sufficient power because if not the data obtained would be useless and hence the animals wasted, something an ACUC is specifically charged to prevent. But the calculation also tells the ACUC the minimum number of animals that would achieve the needed power, and this is used to limit the size of each group.

This is dramatically different to older studies; McCune for instance used 100 mice, and some older vaccine studies in guinea pigs used as many as 400 animals. These animals were obviously outbred and hence the large number would reduce the expected higher variance. However the chance of directly reproducing these studies in the present day are low since ACUCs would not likely permit this.

Where higher numbers of animals are needed are in relapse studies, and our own evaluations of this type of study suggest these studies are under-powered [including our own]. This has not been addressed.

Studies testing new drugs in the mouse and other species take a finite time, raising further ACUC issues. Drugs are usually given by gavage, and while this works well if the technician is gentle, prolonged handling of the mouse can cause be problematic. Esophageal penetration by the gavage needle can cause subcutaneous emphysema and other complications can occur. In studies in which the drug has to be injected [s.c. or i.p.] much shorter periods of time are usually permitted by ACUCs. For long-term studies we and others have tried simply to introduce the drug via the drinking water, but nobody has yet thoroughly evaluated how effective this really is. One can calculate roughly what an animal might drink each day, but then there are extra elements to take into account. The intake may drop because the animal does not like the taste. Also, mice establish pecking orders which might skew intake within a group kept in an individual cage.

Drugs can be toxic, and this is not necessarily noticed initially in small rodents. To address this, some sort of Karnofsky scale should be established, scored by animal husbandry technicians each day. Such scales monitor weight, general behavior, and so forth, but obviously feed and water intake per animal is harder to observe. If animals do show evidence of suffering then most ACUC require immediate euthanasia. The up-side here is that this permits harvesting of organs and blood without any tissue autolysis [or overnight cannibalism by other members of the cage in the case of mice].

Institutions also have Biosafety Committees in the present day. Stocks of M. tuberculosis have to be properly produced, stored in a safe environment, and accounted for. Facilities have to have SOPs in place for worker training and procedures using bacteria. Tuberculosis is a level III agent, and hence has to be kept at all times under these conditions. Facilities should have an Exposure Control plan, as well as emergency procedures. In studies using aerosols, equipment available ranges from whole-body exposure to nose-only. Both are effective, but obviously the instrument must be fully contained and HEPA-filtered. The best location for a level-III facility is stand-alone, but if this is not available and the laboratory is within a building, then it should be on the top floor of that building.

References

1. 1982. [The etiology of tuberculosis by Dr. Robert Koch. From the Berliner Klinische Wochenschrift, Volume 19 (1882)]. Zentralbl Bakteriol Mikrobiol Hyg [A] 251:287-96.
2. Agrawal, S., N. S. Thomas, A. B. Dhanikula, C. L. Kaul, and R. Panchagnula. 2001. Antituberculosis drugs and new drug development. Curr Opin Pulm Med 7:142-7.
3. Almeida, D., E. Nuermberger, S. Tyagi, W. R. Bishai, and J. Grosset. 2007. In vivo validation of the mutant selection window hypothesis with moxifloxacin in a murine model of tuberculosis. Antimicrob Agents Chemother 51:4261-6.
4. Beamer, G. L., and J. Turner. 2005. Murine models of susceptibility to tuberculosis. Arch Immunol Ther Exp (Warsz) 53:469-83.
5. Benedek, T. G. 2004. The history of gold therapy for tuberculosis. J Hist Med Allied Sci 59:50-89.
6. Birath, G. 1969. Introduction of para-amino-salicylic acid and streptomycin in the treatment of tuberculosis. Recollections of the last 25 years. Scand J Respir Dis 50:204-9.
7. Bunn, P., and F. Robinson. 1957. Relative efficacies of chick embryo and standard American Trudeau Society medium in isolation of tubercle bacilli from human sputum. Am Rev Tuberc 76:703-5.
8. Canetti, G., N. Rist, and J. Grosset. 1964. Primary Drug Resistance in Tuberculosis. Am Rev Respir Dis 90:792-9.
9. Collins, F. M. 1993. Tuberculosis: the return of an old enemy. Crit Rev Microbiol 19:1-16.
10. Cynamon, M. H., and S. P. Klemens. 1989. New antimycobacterial agents. Clin Chest Med 10:355-64.
11. Daniel, T. M. 2006. The history of tuberculosis. Respir Med 100:1862-70.
12. Daniel, T. M. 2005. Robert Koch and the pathogenesis of tuberculosis. Int J Tuberc Lung Dis 9:1181-2.
13. Davies, G. R., A. S. Pym, D. A. Mitchison, E. L. Nuermberger, and J. H. Grosset. 2007. Evaluation of new antituberculosis drugs in mouse models. Antimicrob Agents Chemother 51:403; author reply 403-4.
14. Di Pietrantonio, T., and E. Schurr. 2005. Mouse models for the genetic study of tuberculosis susceptibility. Brief Funct Genomic Proteomic 4:277-92.
15. Domagk, G. 1951. The chemotherapy of tuberculosis with thiosemicarbazones. Ir J Med Sci 6:474-85.
16. Domagk, G. 1950. Investigations on the antituberculous activity of the thiosemicarbazones in vitro and in vivo. Am Rev Tuberc 61:8-19.
17. Donoghue, H. D., M. Spigelman, C. L. Greenblatt, G. Lev-Maor, G. K. Bar-Gal, C. Matheson, K. Vernon, A. G. Nerlich, and A. R. Zink. 2004. Tuberculosis: from prehistory to Robert Koch, as revealed by ancient DNA. Lancet Infect Dis 4:584-92.
18. Dubovsky, H. 1973. Koch's remedy for tuberculosis. S Afr Med J 47:1609-14.
19. Ducati, R. G., A. Ruffino-Netto, L. A. Basso, and D. S. Santos. 2006. The resumption of consumption-- a review on tuberculosis. Mem Inst Oswaldo Cruz 101:697-714.
20. Duncan, K. 2004. Identification and validation of novel drug targets in tuberculosis. Curr Pharm Des 10:3185-94.
21. Dunn, P. L., and R. J. North. 1995. Virulence ranking of some Mycobacterium tuberculosis and Mycobacterium bovis strains according to their ability to multiply in the lungs, induce lung pathology, and cause mortality in mice. Infect Immun 63:3428-37.
22. Feldman, W. H. 1969. Tuberculosis chemotherapy: reminiscence of in vivo research. Scand J Respir Dis 50:186-96.
23. Feldman, W. H., A. G. Karlson, and et al. 1949. Parenteral administration of p-aminosalicylic acid in experimental tuberculosis. Mayo Clin Proc 24:220-4.
24. Fitzpatrick, M. 1962. The legacy of Laennec, Villemin, and Trudeau to medicine. Am Rev Respir Dis 86:925-8.
25. Forge, A., and J. Schacht. 2000. Aminoglycoside antibiotics. Audiol Neurootol 5:3-22.
26. Glover, R. E. 2006. A short survey of Sir John M'Fadyean's contributions to the study of tuberculosis. Journal of Comparative Pathology and Therapeutics 1937; 50: 356-76. Vet Microbiol 112:348-68.
27. Goldman, R. C., B. E. Laughon, R. C. Reynolds, J. A. Secrist, 3rd, J. A. Maddry, M. A. Guie, A. C. Poffenberger, C. A. Kwong, and S. Ananthan. 2007. Programs to facilitate tuberculosis drug discovery: the tuberculosis antimicrobial acquisition and coordinating facility. Infect Disord Drug Targets 7:92-104.
28. Grumbach, F., G. Canetti, J. Grosset, and M. le Lirzin. 1967. Late results of long-term intermittent chemotherapy of advanced, murine tuberculosis: limits of the murine model. Tubercle 48:11-26.
29. Gruppo, V., C. M. Johnson, K. S. Marietta, H. Scherman, E. E. Zink, D. C. Crick, L. B. Adams, I. M. Orme, and A. J. Lenaerts. 2006. Rapid microbiologic and pharmacologic evaluation of experimental compounds against Mycobacterium tuberculosis. Antimicrob Agents Chemother 50:1245-50.
30. Gupta, U. D., and V. M. Katoch. 2005. Animal models of tuberculosis. Tuberculosis (Edinb) 85:277-93.
31. Harries, A. D. 2008. Robert Koch and the discovery of the tubercle bacillus: the challenge of HIV and tuberculosis 125 years later. Int J Tuberc Lung Dis 12:241-9.
32. Harvey, A. M. 1977. Tuberculosis: the study of a specific disease at Johns Hopkins. Johns Hopkins Med J 141:198-218.
33. Heifets, L., M. Iseman, and P. Lindholm-Levy. 1988. Application of rapid methods (BACTEC system) in clinical mycobacteriology and in the search for new drugs. Bull Int Union Tuberc Lung Dis 63:19.
34. Heifets, L., J. Simon, and V. Pham. 2005. Capreomycin is active against non-replicating M. tuberculosis. Ann Clin Microbiol Antimicrob 4:6.
35. Herzog, H. 1998. History of tuberculosis. Respiration 65:5-15.
36. Hurt, R. 2004. Tuberculosis sanatorium regimen in the 1940s: a patient's personal diary. J R Soc Med 97:350-3.
37. Iseman, M. D. 2002. Tuberculosis therapy: past, present and future. Eur Respir J Suppl 36:87s-94s.
38. Jain, S. K., S. M. Hernandez-Abanto, Q. J. Cheng, P. Singh, L. H. Ly, L. G. Klinkenberg, N. E. Morrison, P. J. Converse, E. Nuermberger, J. Grosset, D. N. McMurray, P. C. Karakousis, G. Lamichhane, and W. R. Bishai. 2007. Accelerated detection of Mycobacterium tuberculosis genes essential for bacterial survival in guinea pigs, compared with mice. J Infect Dis 195:1634-42.
39. Karlson, A. G., and W. H. Feldman. 1949. The effect of combined therapy with streptomycin and para-aminosalicylic acid on experimental tuberculosis in guinea pigs. Mayo Clin Proc 24:510-5.
40. Kaufmann, S. H. 2005. Robert Koch, the Nobel Prize, and the ongoing threat of tuberculosis. N Engl J Med 353:2423-6.
41. Kaufmann, S. H. 2006. Tuberculosis: back on the immunologists' agenda. Immunity 24:351-7.
42. Kaufmann, S. H., and U. E. Schaible. 2005. 100th anniversary of Robert Koch's Nobel Prize for the discovery of the tubercle bacillus. Trends Microbiol 13:469-75.
43. Khasnobis, S., V. E. Escuyer, and D. Chatterjee. 2002. Emerging therapeutic targets in tuberculosis: post-genomic era. Expert Opin Ther Targets 6:21-40.
44. Kurth, R., and W. H. Haas. 2002. Epidemiology, diagnostic possibilities, and treatment of tuberculosis. Ann Rheum Dis 61 Suppl 2:ii59-61.
45. Lehmann, J. 1964. Twenty Years Afterward Historical Notes on the Discovery of the Antituberculosis Effect of Paraaminosalicylic Acid (Pas) and the First Clinical Trials. Am Rev Respir Dis 90:953-6.
46. Lenaerts, A. J., M. A. Degroote, and I. M. Orme. 2008. Preclinical testing of new drugs for tuberculosis: current challenges. Trends Microbiol 16:48-54.
47. Lenaerts, A. J., V. Gruppo, J. V. Brooks, and I. M. Orme. 2003. Rapid in vivo screening of experimental drugs for tuberculosis using gamma interferon gene-disrupted mice. Antimicrob Agents Chemother 47:783-5.
48. Lenaerts, A. J., V. Gruppo, K. S. Marietta, C. M. Johnson, D. K. Driscoll, N. M. Tompkins, J. D. Rose, R. C. Reynolds, and I. M. Orme. 2005. Preclinical testing of the nitroimidazopyran PA-824 for activity against Mycobacterium tuberculosis in a series of in vitro and in vivo models. Antimicrob Agents Chemother 49:2294-301.
49. Lenaerts, A. J., D. Hoff, S. Aly, S. Ehlers, K. Andries, L. Cantarero, I. M. Orme, and R. J. Basaraba. 2007. Location of persisting mycobacteria in a Guinea pig model of tuberculosis revealed by r207910. Antimicrob Agents Chemother 51:3338-45.
50. Lenaerts, A. J., C. M. Johnson, K. S. Marrieta, V. Gruppo, and I. M. Orme. 2005. Significant increases in the levels of liver enzymes in mice treated with anti-tuberculosis drugs. Int J Antimicrob Agents 26:152-8.