Missing Microbes: How the Overuse of Antibiotics Is Fueling Our Modern Plagues (23 page)

In March 2011, Laurie began the DuraSTAT experiments, so named because we were testing the durability of a brief antibiotic exposure to produce an effect. She divided the mice into four groups: no antibiotics, which was the control group; STAT for only four weeks and then stop; STAT for eight weeks and then stop; or STAT for the duration of the experiment. All of the mice were put on a high-fat diet at six weeks to bring out any differences. Laurie focused on females because of the results of our FatSTAT study.

Mice getting continuous antibiotics for the duration of the experiment gained weight compared to the controls, just as expected. But the effects of getting antibiotics for four weeks or eight weeks were the same as for twenty-eight weeks. The mice receiving the penicillin gained 10–15 percent more in total weight and 30–60 percent in fat compared to the antibiotic-free control mice. In other words, exposure to STAT early in life was sufficient for a lifelong effect; the development of the mice changed. Although the results in DuraSTAT were not identical to those of FatSTAT, neither were the experimental conditions. So the experiments are not directly comparable. The relevant comparisons are within each experiment. This is an important issue in science, where investigators have gone astray by comparing the effects in one experiment with those in a different one; conditions change in ways that often are not being measured. But for us, the trends were exactly the same: early-life STAT permanently changed development in these mice.

Next we decided to study the microbiome itself. Laurie had been faithfully collecting the tiny fecal pellets, often once a day, from every mouse. She had thousands of little plastic test tubes in white boxes, one pellet per tube, one hundred tubes per box. It would take about eighteen thousand pellets to make a pound. They were worth more than their weight in gold because of the secrets they carried.

Laurie sequenced hundreds of specimens to determine their DNA compositions and to learn about the structure of their microbial communities (let’s say again assessing the ratio of teachers to police officers but now in much greater detail), including tax lawyers, taxi drivers, and taxidermists.

First she looked at samples from newly weaned three-week-old mice that had been given penicillin and compared them with samples from control mice that did not get the drug. Although the community structures of the two groups overlapped some, they were clearly different. This was exactly as we expected: antibiotics affect the structure of the microbial community in the intestinal tract.

Then we looked at pellets obtained eight weeks into the study. Now there were essentially three groups of mice: the controls (no antibiotics), the mice still receiving antibiotics, and those who stopped antibiotics after four weeks and drank plain water for the next four weeks. As expected, the microbial community structure of controls and mice on continuous antibiotic were even more dissimilar than they were at three weeks. Antibiotics work. But the microbial community structures of the mice that had stopped taking antibiotics now looked just like those of the controls; they nearly overlapped. This means that the major effects of four weeks of antibiotics on the community structure were just transient. This was very clear. Yet, remember, these mice got just as fat as the others, which suggests that a brief exposure to antibiotics early in life, which causes an early perturbation of resident microbes, can lead to a lifelong effect. And the perturbation need not be permanent.

This is a key finding. I believe that it is the paradigm for what is happening to our children. Disturbing the microbes of mice during this critical early window is sufficient to change the course of their development. This was the experiment that proved to me that antibiotics have the potential to change development. And, of course, development is multidimensional: it is metabolic, as we were studying in the mice, but it is also immunologic and cognitive. As babies grow, while they are sleeping and dreaming, the context of their later development is being formed in partnership with their ancient microbes. Even transient perturbations at that critical time can make a big difference.

But we are scientists, and we need to keep extending the story, learning the details, and finding the mechanisms. We need to answer the seemingly simple question: How does it work? What is so important about antibiotic exposure? Is it just its effects on the microbes, or is it because the penicillin had other effects on the body, directly interacting with the tissues of the mouse, irrespective of the effects on the microbes? As with many prior experiments, including those conducted by Jeff Gordon, we would attempt to answer the question by transferring microbes between mice.

Recall our earlier question: Was weight gain a direct effect of the antibiotics or was it a result of how the antibiotics affected resident microbes? We presumed it was the microbes, but presumption is not proof. To find out, we needed to transfer the STAT or the control microbes into a neutral situation and then observe whether there were differences in the recipients. Like Jeff, we chose to study the effects in germ-free mice.

We bought fifteen germ-free female mice, and in late August 2011 they arrived in three plastic bubbles, five to a bubble, newly weaned at three weeks old. The company told us that we could maintain them in the bubble for up to seventy-two hours, enough time to start our experiment. We called it TransSTAT because we were transferring the STAT-affected microbiota to recipient mice.

Laurie chose six eighteen-week-old mice from her DuraSTAT experiment: three controls and three on continuous antibiotics. She collected cecal contents from each mouse and pooled them into two groups, one from controls, one from STAT mice. Calling on her extensive background in bacteriology, Laurie took special steps to preserve the viability of the microbes, some of which are so sensitive to oxygen that even a brief exposure to air kills them. Then she introduced the cecal matter into the stomach of each germ-free mouse. Seven received pooled cecal contents from controls, and eight received the cecal contents from STAT mice. To you and me the introduction of cecal contents into the stomach seems particularly unappetizing, but mice are coprophagic, meaning that they regularly eat their own feces as well as the feces of cohoused mice.

Now the mice were no longer germ-free. They had been “conventionalized” and could begin the next phase of their lives with their own residential microbes. We followed them for five weeks, obtaining frequent fecal samples and taking measurements, including DEXA scans, four times on each mouse. None of the mice received any antibiotics. All were raised identically, differing only in which microbes they received.

As expected, all of the mice gained weight since they were still growing. However, the mice given the STAT microbes gained more weight and had more fat than the mice that were fed the control microbes. Nor were the effects small. The STAT recipients gained about 10 percent more weight and about 40 percent more fat than the control recipients did.

With this experiment, Laurie proved that the STAT-induced changes in development were transferrable by altered microbes alone.

*   *   *

STAT showed us what happens on the farm. But I am mostly interested in human children. When they get antibiotics, the dosage is rarely continuous. Rather, as discussed earlier, they receive short courses, usually five to ten days, depending on the problem (ear infection, bronchitis, sore throat) and on the doctor.

I wanted to see whether short pulses of antibiotics would affect weight gain and fat. Thus came the new model, which we called PAT, for pulsed antibiotic treatment. Instead of low doses, mice got the antibiotics just like children do, full therapeutic doses for just a few days in several pulses.

We chose amoxicillin and tylosin, which together represent more than 80 percent of all the antibiotics prescribed for American children. We then selected four groups of mice: controls that did not get antibiotics, a group given amoxicillin in three pulses, a group given tylosin in three pulses, and—thinking there might be an additive effect—a mixed group that alternated tylosin, amoxicillin, tylosin for their three pulses.

To get the drugs into the baby mice as early as possible, Yael bred adult females and put antibiotics (or not, for the controls) into their water ten days after they gave birth. We guessed the drugs would be absorbed into the bloodstreams of the moms and get into their milk and thus affect the microbes in their pups—an assumption that proved to be correct.

The first PAT exposure occurred when the babies were ten to fourteen days old. At twenty-eight days old, after they were weaned and on their own, and again at thirty-seven days old, they got three-day pulses in their drinking water. On day forty-one, we switched all the mice to a high-fat diet, so we could enhance differences induced by antibiotics. All the mice we studied were females, because the breeding resulted in a more regular group than the males.

By day twenty-eight, all the PAT mice were growing significantly faster than the controls. We performed analyses on fat, bone, and muscle for the next 150 days of life, which took the mice well into middle age and early old age. PAT mice showed more muscle mass than the controls but not much difference in fat mass. Bone was a different matter. The PAT mice that received amoxicillin showed increased bone area and mineral content for the duration of the experiment. Perhaps the effect was permanent because they received the drugs so early in life. And since amoxicillin is the most frequently prescribed drug in childhood, I can only wonder if that’s the drug that most promotes the recent increases in human height.

Yael had collected more than three thousand fecal pellets from the mice and, for each specimen, knew which mouse it came from, on which day, and which treatment the animal got. With help from colleagues at Washington University in St. Louis we took a close look at their DNA. We wanted to know how our treatments had affected each animal’s intestinal microbial diversity.

We found that moms had an average of 800 species in their fecal pellets. After one pulse, pups in the control group were like the moms. Those in the amoxicillin group had about 700 species. But mice in the tylosin and mixture group had only 200 species. In other words, the one course of antibiotics had caused the suppression or disappearance of about two-thirds of the usual bacteria in their feces. We saw a similar effect with amoxicillin, but it was much milder.

Now, after the three courses were over, we wondered whether the richness and biodiversity of the bacterial species would bounce back. It mostly did for amoxicillin, which is a relatively mild drug. But in the mice that had received the tylosin, the diversity never went back to normal, even months after the last antibiotic dose. Tylosin had permanently suppressed or wiped out a proportion of the organisms passed on to them by their mother.

We also measured the so-called evenness of microbial diversity. If it’s high, it means most species are found in roughly equivalent numbers. If low, only one or a few species dominate. In a human society, we could compare peacetime, when many different professions are well represented, and wartime, when there is a huge increase in the number of soldiers and corresponding decreases in all other professional groups. In war, the professional structure of society changes markedly. Tylosin treatment gave us the wartime equivalent with low evenness. PAT was causing permanent changes to the structure of the microbial community early in life, just as the mice were developing.

*   *   *

All told, our STAT and PAT experiments built a strong story that early-life antibiotics change the development of mice through their effects on resident microbes. But mice are not humans. We wanted to know if anyone was trying to link obesity with antibiotic use in young children. Despite a profusion of published studies on childhood obesity—including investigations into birth weight, time spent watching TV, amount of exercise, exact details on every dietary nuance—as well as some big studies now under way, no one to our knowledge had ever asked about antibiotics.

Then my colleagues Drs. Leo Trasande and Jan Blustein heard about the Avon Longitudinal Study of Parents and Children (ALSPAC) study in Britain. Beginning in 1991 more than 14,500 pregnant women in the Avon Health District were recruited for a study. Their children, enrolled at birth, became a cohort that was studied for the next fifteen years. We were particularly interested in the kids who became overweight or obese.

Luckily for us, there was exactly one question on the questionnaire each parent filled out periodically. As part of a survey of the drugs to which the kids were exposed, they asked: “Did your child use any antibiotics in the preceding period?” It was asked when the kids were six, fifteen, and twenty-four months of age.

Nearly a third had received antibiotics in the first six months of life. By age two, three-quarters had been treated. Did the antibiotics make any difference? The calculations were complex, and it took an excellent statistician like Leo to make sense of them. Leo had to examine the effects of antibiotics while controlling for such factors as the baby’s initial weight and the mom’s weight and whether the baby had been breast- or bottle-fed and for how long.

The upshot: children who received antibiotics in the first six months of life became fatter. We weren’t surprised; the earlier in life, the stronger the effect on farm animals. Laurie had shown that early-life dosing was more important in the mice and that if we had to guess which time period would be most important for the development of human babies, it would be the first months of life.

So on the farm, in our mouse experiments, and in an epidemiologic study of human children, there was consistent evidence that early-life exposure to antibiotics could change development leading to larger size and more fat. We are testing more variations in mice, but the story continues to hang true, as we begin to fill in more and more details of the plot and the characters.

*   *   *

After the first epidemiologic study, Jan and Leo used ALSPAC to look at methods of birth delivery. Using parallel statistical analyses, they found that C-section births also were associated more with obesity. This was one of several studies concerning U.S., Canadian, Brazilian, and now English children published in 2011–13 that examined the same question. There were differences in design and findings among all of the studies—for example, we found that nearly all of the effect occurred when the mom already was overweight. However, in each studied population C-section was associated with worse outcomes but the other (confounding) factors may also contribute to the risk. No one had ever made the connection between C-sections and childhood obesity. Maybe the informed consent form that a woman signs before she undergoes a C-section will say in the future: “One of the risks of this procedure is that your baby has an increased risk of becoming obese, and developing celiac disease, asthma, allergies…”