In spite of the hassles with contamination, generating the amino acid samples turned out to be the easiest experimental task by far. The real trick, mastered by only a handful of analytical chemists in the world, was to determine whether the calcite crystal faces had preferentially selected D- or L-aspartic acid.
Tim Filley is an expert at organic chemical analysis, so we were confident that we could do the work ourselves. We had an adequate gas chromatograph/mass spectrometer, the analytical machine of choice; we had a chiral chromatographic column that should separate D- from L-aspartic acid; and we had a well-documented procedure for preparing our samples prior to injecting them into the GCMS. The setup took some time, but the methods seemed straightforward.
After a lot of practice with aspartic acid standards in different D:L ratios and different concentrations, we were ready to try the real thing. Our first analyses took place just before Christmas 1999.
GCMS affords a kind of rapid (if not instantaneous) gratification, because the ratio of two compounds, such as D- and L-aspartic acid, corresponds to the areas of two peaks that appear side by side on a graph. Joined by Tim's wife, Rose, who is also a skilled analytical chemist, we injected our first sample and waited. It took the better part of 20 minutes for the molecules to work their way through the 50-meter-long coiled chiral column, which was packed with chemicals that slow down L-amino acids slightly more than D. D-aspartic acid passed through the column about 20 seconds faster than L-aspartic acid. The first peak showed up sharp and clear—plenty of D-aspartic acid molecules there. The next 20 seconds seemed to take a lot longer than any 20 normal seconds should, but the second peak, representing L-aspartic acid, appeared on schedule. And it was clearly bigger—at a guess, more than 10 percent bigger. We allowed the machine to run just a few minutes longer, to make sure there weren't any obvious contaminants, and then stopped to let the computer calculate the peak areas.
The actual ratio turned out to be closer to a 5-percent excess of L, but it was a tantalizing result that clearly pointed to chiral selection. We decided to run the same sample a second time, just to be sure. Twenty minutes later, the pair of peaks appeared again, but this time we saw no more than a 1-percent excess of L, not enough to be sure the effect was real. The average value was a 3-percent excess, while our experimental error was evidently close to ±2 percent. Tim suggested that we run a couple of more standards to be sure the machine was giving us clean peaks.
We were off and running, with 22 more samples, each to be prepared and run in duplicate, plus standards after every few sample injections. Given the prep time, it took more than two hours to test each sample or standard. We started running in three overlapping shifts. With two small girls, Rose was always up early and did the first few injections. Tim and I usually couldn't stay away, so we'd take over in the morning and keep things running through the evening and into the wee hours. For the first couple of days, we were there 18 or more hours at a stretch, running sample after sample, eagerly watching the screen as the pair of peaks appeared 20 minutes into each run. It was addictive and exhausting, like a slow-motion video game.
So it went: Run a few samples with tantalizing results, then more standards. One encouraging sample suggested a 4-percent excess of D-aspartic acid—a critical result, since L contamination, the life sign, was everywhere, but an excess of D could come about only through crystal selection. Still, the results weren't consistent. What was even worse, the peak shapes began to broaden and degrade after only about 25 injections. Broad peaks overlap, making analysis of the D:L ratio all but impossible.
We called the supplier of the chiral column to complain—a new $500 column like ours should have lasted for hundreds of runs, not just 25. They suggested that we bake out the column to make sure that large, unwanted molecules hadn't contaminated it. We had been turning off the machine a few minutes after we saw the aspartic acid peaks, because we were eager to do the analyses as fast as possible. Perhaps other slower molecules hadn't made it all the way through the column and were gumming up the system.
After the prescribed treatment, the column worked better for a few injections, but soon the peak shapes were worse than ever. We called the company again.
“What kind of samples are you running?” they asked.
“Just aspartic acid,” Tim said.
“Are you sure there's nothing else in the samples?” they responded.
“Well, maybe a little bit of dissolved calcite; it's just calcium carbonate.”
That was it. The company rep told us that the mineral residue had quickly degraded our column, making it useless for further work. They generously offered to send us another for free (probably suspecting that we'd be going through a great many $500 columns in the coming months).
So now we'd have to execute an additional chemical step to remove the calcium carbonate with acid before injection. We weren't happy about that, because each additional treatment of the samples increased the chance of contamination.
We kept trying, but those first experiments were just too flaky. Lots of L excesses of a few percent, one or two D excesses, but nothing really systematic. What's more, some of the L excesses came from the fracture surfaces, which should have had no effect, while the right- and left-handed calcite faces gave inconsistent results at best.
We had spent months at the project with nothing to show for it but a lot of experience under our belts. In the spring of 2000, Tim's postdoc was nearly up. He and Rose were looking forward to setting up a new home in Indiana. I had to make a decision whether, and if so how, to continue.
I decided to give it another shot. One thing was absolutely clear: The experiments had to be performed in ultraclean facilities, with every possible precaution against contamination. My colleagues at our sister lab, the Department of Terrestrial Magnetism (DTM for short), maintain a chemical clean lab at our Northwest Washington campus for isotope and trace-element work, and they generously let me use one of their high-tech chemical hoods for a few months.
In May of 2000, I hauled box upon box of chemicals, glassware, and crystals to the second-floor clean lab of DTM's venerable experiment building. The sparkling, well-lit labs could be entered only through a small glassed-in vestibule, where the transition to clean-room protocols took place. I removed my shoes and slipped into a pair of pale-blue booties, white coveralls, hood, respirator mask, and latex gloves. The protective gear wasn't uncomfortable, but it took a bit of getting used to.
After organizing my supplies, I began the experiment. Leaving nothing to chance, I treated the thrice-washed crystals with extra care, used the highest-purity amino acids, ultra-pure water (at $40 per gallon!), and freshly baked glassware. Every step was carried out in a chemical hood, a glass-enclosed volume about 4 feet wide, 2 feet deep, and 3 feet high—plenty of room for my beakers. My hood was maintained with positive pressure to prevent outside air from entering the enclosed area.
The experiments were conducted almost exactly as before. First, soak four crystals for 24 hours in the 50:50 amino acid bath. This time I made sure that the pH of the solution remained close to 8, a value ensuring that the calcite would not start to dissolve. Then, after a day of soaking, I repeated the now-familiar wash process, applying acid to each crystal, face by face. The day's work produced 23 small glass vials of acid extract. Over the next two days, I repeated the entire experiment two more times to be absolutely certain.
A week's work yielded 69 vials of aspartic acid solution washed from crystal faces, plus several vials with individual samples of each day's aspartic acid solution, of the pure water, and of the acid wash. I wanted samples of everything, in case I had to track down contaminants. More than 75 sample vials needed to be processed, each to be analyzed at least 3 times. Two-hundred-plus amino acid analyses is a huge job, and the Geophysical Lab facilities were not up to it. So for that crucial, final step I headed downtown, hat in hand, to George Washington University, to see geochemist Glenn Goodfriend, a former colleague at the Geophysical Labor
atory.
Success in a scientific career can be measured in many ways. Some scientists crave admiration and respect from their peers. Others prize a flexible job that gives them the freedom to pursue any line of research. And for many researchers, the quiet exuberance of doing good science is the prime measure of a happy and successful career. By all these measures, Goodfriend was one of the most successful scientists I'd ever known.
At the agreed-upon late-morning hour, I arrived at his modest office in the basement of Lisner Hall, home of the Geology Department. Glenn was a research professor, his work sustained almost entirely by a succession of two- and three-year grants from the National Science Foundation. Few scientists have the stature and stamina to survive like that for long, but Glenn was a master with more than a decade of steady funding.
“Hi, Bob!” he said, with his usual big smile, his thick black mustache and curly black hair making him seem younger than his years. (“It comes from drinking lots of good red wine,” he'd say.) Piles of manuscripts and journals covered most of his desk and surrounding tables; banks of neatly labeled filing cabinets hinted that he had the upper hand on entropy. Glenn leaned back in his chair, hands behind his head—a characteristic gesture I'd soon come to learn. “What's up?”
I described the chirality experiments and their implications for origins research in as sexy a way as I could. Glenn nodded often, but his smile slowly faded. When I had finished, he launched into an intimidating list of his own amino acid projects already underway.
Glenn's research exploited the fact that although almost all of life's amino acids are left-handed, as soon as an organism dies, a slow, inexorable process called racemization—the random flipping of molecules from L to D and vice versa—begins. Eventually, after a few tens of thousands of years, an organism's amino acids will have completely randomized to a 50:50 mixture. This tendency for the D:L ratio of amino acids to change over time provides a powerful dating technique: The older the shell or bone, the closer its amino acids will be to a 50:50 mix. Other factors—notably the average water temperature, the acidity, and the salinity—also affect the rate of racemization; the D:L ratio in a fossil can thus provide evidence for changes in ancient environments. Glenn was one of the world's experts in determining that crucial ratio, so scientists from all over asked for his help. In one ongoing collaboration, he determined the ages of fossil eggshells from Australia, to help understand long-term changes in the continent's vegetation. Another project used clam shells to measure recent changes in the salinity of the Venetian Lagoon. He also was studying amino acids in fossil shells from the Baja Peninsula of Mexico to deduce patterns of climate change.
But his biggest and boldest effort was his long-term collaboration with Harvard paleontologist Stephen Jay Gould on the evolution of Cerion, a beautiful little Bahamian land snail. Glenn had helped to collect countless thousands of these inch-long shells from deep pits dug into remote sand dunes. The major part of the collection was exhumed from a deserted stretch of Long Island in the Bahamas. Glenn's Cerion specimens displayed remarkable variations, even though all were members of a single species. Some shells were elongated, while others were almost round; some richly decorated, others almost smooth. These and several dozen other morphological characteristics provided Gould with a perfect species to test his provocative theories of evolutionary change. Glenn's job was to provide the critical dating by analyzing D:L ratios from thousands of individual shells. Once supplied with enough differently shaped shells, their ages, and the DNA analyses performed by another colleague, Gould hoped to tease out the evolutionary pathways of gradual morphological change. Years, maybe decades, of work lay ahead.
Given these commitments, Glenn was certainly too busy to take on a new project. Yet he was also intrigued. Chiral selection was a new challenge for his analytical system, and he knew a good project when he saw it.
“Looks like it's about time for lunch!” he said, abruptly changing the subject.
“Let me take you.” I sensed a setup, but I would have done just about anything to secure his help.
“There's a nice little place a couple of blocks from here. Kinkead's.” It wasn't a question.
“Sure, let's go.”
Kinkead's specializes in seafood, to which Glenn was deathly allergic; he even had to wear protective gloves when handling his favorite Cerion shells. But Kinkead's had a great wine list, and Glenn had a passion for good red wines. Glenn ordered glasses of two different wines and extra glasses for each of us, so we could compare and contrast. Some months later, I learned that Kinkead's was a kind of test; had I balked at the noontime diversion, our collaboration might never have happened.
Evidently I passed. “OK,” he said, and paused. “You'll have to derivatize the samples, but I'll do the analyses.” So I would have to do a bit of chemical prep work, but I was in business.
Glenn had to maintain his analytical facility in the same large room as the undergraduate anthropology lab at George Washington. The first thing you notice on entering is the inordinate number of bones—dozens of human skulls, legs, ribs, and hip bones in wooden trays and glass display cases. A fully mounted human skeleton (lacking only the odd digit and forelimb) presides slack-jawed over the unsettling scene. A long, black-topped table surrounded by two dozen padded stools occupied the center of the 25 × 40-foot space. Glenn had a cramped 3-foot-square chemical hood on one side of the room and his arsenal of state-of-the-art gas chromatographs along the opposite wall, where any undergraduate might inadvertently bump into them. How could anyone work effectively under these conditions, I wondered? And yet one quickly learns to focus only on the diminutive vials and their secrets.
I showed up there mid-morning of the following week to prepare my amino acid samples for analysis. The aspartic acid had to be chemically modified so that the D- and L-amino acids could be separated more efficiently by gas chromatography. The amino acid molecules, which normally dry to a white powder, had to be treated so that they evaporated to a gas at high temperature. Under Glenn's guidance, I made sure each sample was completely dried down, then added a milliliter of thionyl chloride, an orange-tinged toxic liquid, and tightly capped the vials under a stream of nitrogen gas. Then we cooked the samples, two dozen at a time, in the oven.
I have never watched a scientist more meticulous in his procedures than Glenn, who proved to be one of the most exacting, finicky experimentalists I'd ever met. Like a master chef, he prepared amino acid samples for analysis the same way every time. He heated them at 100°C for one hour in a small, squat oven, instructing me to open the oven door quickly and place the tray of vials on the shelf in one swift gesture. Close the door within 4 seconds to keep the temperature at the proper level. If the temperature dropped even 0.2°C, he recorded it in his lab notebook.
An hour later, to the second, I had to remove the samples from the oven with a similar smooth motion. If I was 10 or 15 seconds late, his mustache would twitch and the discrepancy went into the notebook. One secret to Glenn's success was his absolute, rigorous reproducible procedures.
Once the rows of vials had cooled, I opened each one, dried them under a vacuum, added a second chemical (trifluoroacetic acid anhydride), sealed the vials, and heated them again for exactly five minutes. At the end of this process, each amino acid sample had been modified to a volatile form that was ready to analyze. We transferred a small volume from each into glass autosampler vials, loaded up the gas chromatograph, and set it to run overnight.
Glenn and I were paranoid about the potential for unconscious bias. We knew exactly the chiral effects we were looking for—certain faces should select L-molecules and others D-molecules, while the fracture surfaces should display no preference. So I randomly renumbered the samples and Glenn renumbered them again in his own notebook. That way, neither of us would know which sample came from which face until after we'd completed all the analyses and compared numbers. It's all too easy to see what you want to see in random data. Once the samples were prepared
, we had only to wait for the automated machine to do the analyses. Glenn promised to call me the next day with our first results.
“Hi, Bob. Looks like we have some data,” he reported the next afternoon. “Got a pen?” I scribbled down a long list of specimen numbers and D:L ratios. Quite a few of the numbers were close to 1.00—no effect. But there were also several values significantly higher and lower: 0.958, 1.031, 0.965, and other numbers that pointed to a possible chiral effect.
“Of course we'll have to repeat all these analyses a couple more times,” Glenn added. I was to learn that performing analyses in triplicate (at a minimum) was one of his trademarks.
“What sorts of reproducibility do we have?”
“Looks like about plus-or-minus half a percent. Not bad.” I was amazed. Errors smaller than 1 percent were almost unheard of in this business.
As soon as I had sorted out which analysis went with which face, a clear and compelling story began to emerge from the data. Left-handed calcite faces almost universally displayed D:L ratios a few percent less than 1.00. These faces preferentially retained L-aspartic acid. The right-handed calcite crystal faces displayed an equal and opposite affinity for D-aspartic acid. Equally important, all of the nonchiral fracture surfaces, which served as our experiment's internal control yielded D:L ratios indistinguishable from 1.00. Glenn's repeat analyses of each sample over the next week reinforced the story.
We wrote up the results quickly and submitted the short manuscript to the Proceedings of the National Academy of Sciences, with Hat Yoder serving as the sponsoring Academy member. The discovery that chiral crystal faces of calcite selectively adsorb D- and L-amino acids suggested not only a plausible chiral environment on the early Earth, but also a possible mechanism for making functional biological macromolecules. If adsorbed L-amino acids lined up sequentially on the crystal surface, then they might be poised to link to each other, forming a protein-like polymer of amino acids. Perhaps in this way mineral surfaces selected, organized, and assembled the first homochiral biomolecules.
Genesis: The Scientific Quest for Life's Origin Page 21