New & Noteworthy
January 23, 2014
Ponce de Leon searched the New World for the fountain of youth. Turns out that if he had some of the tools at our disposal, he wouldn’t have even had to leave Europe. He just needed to go to the local bakery or brewery and look inside the yeast he found there. Of course, then he wouldn’t have found Florida…
Using in silico genome-scale metabolic models (GSMMs) in yeast, Yizhak and coworkers identified GRE3 and ADH2 as two genes that significantly increased the lifespan of yeast when knocked out. Even more importantly, their method also allowed them to identify the mechanism behind this increased lifespan—the mild stress of increased reactive oxygen species (ROS). This last finding may help scientists identify drug targets that they can target to increase the lifespan of people too. If only Ponce de Leon had lots of -omics data and a powerful computer or two!
After constructing an in silico starting state, Yizhak and coworkers entered two sets of data from previous work that had been done on aging in yeast. They next used gene expression profiling to identify which metabolic reactions were different and which were the same in young and old yeast. They then systematically tested the effect of knocking out these reactions one at a time in their computer model to identify those that could potentially transform yeast from old to young with minimal side effects.
Their first finding was that many of their best hits, like HXK2, TGL3, and FCY2, had already been identified as important in prolonging a yeast cell’s life. They decided to look at seven genes that had not been previously identified as being involved in aging.
When two of these seven, GRE3 and ADH2, were knocked out, these yeast strains lived significantly longer with minimal side effects. For example, the strain lacking GRE3 lived ~100% longer than the wild type strain.
Figuring out why these yeast probably lived longer was made simpler because they used metabolic models to identify the genes. The hormesis model of aging suggests that mild stress, like that found in caloric restriction, can lead to increased life span. With this model in mind, the authors focused in on the possibility that knocking out GRE3 and/or ADH2 would lead to increased stress through the production of increased levels of ROS. When they looked, they found that the two knockout strains did indeed have higher levels of two common forms of ROS, hydrogen peroxide and superoxide.
Of course none of us is particularly interested in extending the life of a yeast! But these results could suggest new drug targets to go after that might mimic the effects of caloric restriction without us having to starve ourselves. And these same methods can be used on human cells to find key pathways to target in people. In fact, the authors have started to use their computer models to investigate aging in human muscle cells and found that like in yeast, many of the genes they have identified are consistent with previous work on human aging.
Now we probably shouldn’t get too far ahead of ourselves here. This is a promising first step but it really isn’t much more than Ponce de Leon boarding his ship to begin his trip to the New World. We still have a long voyage ahead of us before we find the fabled fountain of youth.
January 16, 2014
We all know that potato chips are delicious. But we also know that eating too many of them isn’t very good for our arteries or our waistlines. And apparently these aren’t the only chips that can be too much of a good thing.
Chromatin immunoprecipitation (ChIP) is an incredibly valuable technique that lets us see where a particular protein binds in a genome. It can show us the target genes of a particular transcription factor, the distribution of RNA polymerases as they transcribe genes, the places where silencing proteins bind to turn off expression of particular regions, and lots more.
But just like potato chips, more ChIP results aren’t always better. Teytelman and coworkers, publishing in Proceedings of the National Academy of Sciences, and Park and coworkers, publishing in PLoS ONE, have discovered that highly transcribed regions of the genome consistently give false positive ChIP results. In other words, very active regions of the genome look like everything is binding there even when it almost certainly is not. Teytelman and colleagues call these regions “hyper-ChIPable”.
Far from being a reason to despair, though, the discovery of this artifact explains some puzzling previous results and inspires the creation of new, more reliable ChIP methods. This is exactly what Kasinathan and coworkers have done, in a recently published paper in Nature Methods.
The idea behind the ChIP technique is that if you want to know all of the places across the genome where your protein of interest binds, you can lyse cells, shear the DNA into relatively short fragments, and immunoprecipitate your protein from the mixture. Usually the protein and DNA are cross-linked before immunoprecipitation, to strengthen their bond during the rest of the procedure.
After immunoprecipitation, the DNA fragments associated with the protein can be identified using a variety of methods. Finally, mapping the sequences of the fragments to the genomic sequence shows us all the sites that the protein occupies.
Teytelman and colleagues used ChIP-seq to ask whether the silencing complex (Sir2p, Sir3p, and Sir4p) ever binds to non-silenced regions of the genome. They thought they might see some binding, but they were astounded to find significant binding of the complex at 238 distinct euchromatic (non-silenced) loci. This didn’t really make sense, since the yeast Sir proteins are extremely well-studied and there were no biological hints that they have such a large presence at non-silenced genes.
As a control, they looked at previously published ChIP data on the locations of two unrelated proteins, Ste12p and Cse4p, and found that their binding was enriched at the same 238 loci. Finally, they did a ChIP study using green fluorescent protein (GFP) alone. Sure enough, the ChIP data showed that this jellyfish protein apparently bound strongly to chromatin at those 238 sites! The common denominator shared by these loci: they were all very highly expressed.
Meanwhile, Park and coworkers were embarking on a similar journey. They found using ChIP-seq that several unrelated transcription factors seemed to have common targets, which didn’t make biological sense. Control experiments looking at binding sites of Mnn10p (a cytoplasmic protein not expected to have any contact with DNA), or even using nonspecific antibodies that didn’t recognize any yeast proteins, still gave the same set of ChIP targets. Again, these targets were all highly expressed genes.
Each group found several factors contributing to this artifact, although all the reasons why highly expressed regions yield false positives may not yet be uncovered. But whatever the reasons, this finding helps explain some previously perplexing results – such as binding of Mediator complex all over the genome, or the paradoxical binding of silencing regulator Sir3p to the GAL1-GAL10 regulatory region under conditions where transcription is activated, not silenced.
In response to these issues, many researchers are actively trying to improve the ChIP technique. Kasinathan and colleagues have devised a method that they call ORGANIC (Occupied Regions of Genomes from Affinity-purified Naturally Isolated Chromatin) that eliminates crosslinking and substitutes micrococcal nuclease treatment for sonication (to shorten the DNA fragments). In a pilot project, they mapped binding sites for the transcription factors Reb1p and Abf1p. The method looks to be both accurate and sensitive. Most binding locations that they found contained the binding motif sequence for that transcription factor, and also correlated with in vivo occupancy as determined by Dnase I footprinting – both of which support their biological relevance. Importantly, the technique shows no bias towards highly expressed regions.
The lesson for researchers is that ChIP results for highly expressed genes, particularly those done using older protocols, need to be viewed cautiously. And of course this artifact could be an issue for organisms other than yeast. ChIP experiments are used across species, and have been valuable in elucidating the targets of disease-related proteins like the tumor suppressor p53.
The fact that yeast genetics and molecular biology have so well established the roles of certain chromatin-associated proteins was a key part of this puzzle, helping to point out the artifactual nature of some of the ChIP results. Just as a new recipe for potato chips could allow us to eat more of them while staying healthy, yeast research has led the way to a new recipe for more accurate ChIP studies.
Aside from the molecular biology behind this work, it is quite interesting from a sociological point of view as well. What is it like to make a discovery that calls into question a routinely-used technique and a lot of published results? Lenny Teytelman’s blog post on this topic provides a fascinating glimpse into this situation.
January 8, 2014
The janitor on the U.S. comedy series Scrubs is always coming up with terrible inventions. One of his worst was the knife-wrench. It is what it sounds like—a tool with a knife at one end and a wrench at the other.
Of course not all dual purpose tools have to be so useless. Imagine a tool like the one at the right with a razor at one end and a toothbrush on the other. Now you can easily brush your teeth and shave in the shower or at your bathroom sink (as long as you are careful not to cut your cheek).
Turns out that biology has these dual purpose tools too except that they are almost always more useful. For example, Lahudkar and coworkers show in the most recent issue of GENETICS that Cet1p doesn’t just help out with capping mRNA. No, these authors found that it also helps clear RNA polymerase II (RNA pol II) away from promoters. And what’s most interesting is that this second function has little to do with its job in mRNA capping.
Basically the two functions are probably in the same protein because they both happen in the same place, at the start site of a promoter. Just like our brazor is useful because both jobs happen in the bathroom.
The first step was to show that in the absence of Cet1p, RNA pol II was more likely to be found near the start of transcription. The authors showed that this was the case by using a temperature sensitive mutant of Cet1p and a chromatin immunoprecipitation (ChIP) assay targeted at RNA pol II—there was more RNA pol II crowded near the promoter at the nonpermissive temperature.
The next set of experiments showed that merely messing with the cap is not sufficient to cause the polymerase to pause. Lahudkar and coworkers found that RNA pol II occupancy was unchanged in strains carrying mutations in STO1 (also known as CBP80) or CEG1, two components of the capping machinery. Cet1p apparently has a separate, unrelated function in helping to clear polymerases away from the start site of transcription.
The final set of experiments showed that the unpausing activity of Cet1p was found in a different part of the protein from its capping function. Cet1p be can be broadly divided into three regions—a poorly characterized N-terminal domain (amino acids 1-204), a Ceg1p interaction domain (aa 205-266), and a triphosphatase domain (aa 265-549). The last two domains are critical to its capping function.
Lahudkar and coworkers found that deleting the 1-204 aa domain from Cet1p caused polymerase stalling at the promoter without affecting its capping ability. And conversely, that when they impaired the ability of Cet1p to perform its capping function while retaining its 1-204 aa domain, RNA pol II escaped the promoter at the same rate as it did in the presence of wild type Cet1p. A final experiment showed that just expressing the first 300 amino acids of Cet1p was sufficient to get the polymerases moving.
All in all these experiments provide strong evidence that Cet1p has two separate functions—an enzymatic role in capping mRNA and an unrelated activity that helps clear RNA pol II from the regions around the promoters of genes. Which all goes to show that even when you think you have a handle on a protein, it can still surprise you with something new. Turn it around and you just might find a toothbrush at the end.
December 19, 2013
Just like the chicken or milk you buy at a store, chromosomes have a shelf life too. Of course, chromosomes don’t spoil because of growing bacteria. Instead, they go bad because they lose a little of the telomeres at their ends each time they are copied. Once these telomeres get too short, the chromosome stops working and the cell dies.
Turns out food and chromosomes have another thing in common—the rates of spoilage of both can be affected by their environment. For example, we all know that chicken will last longer if you store it in a refrigerator and that it will go bad sooner if you leave it out on the counter on a hot day. In a new study out in PLoS Genetics, Romano and coworkers show a variety of ways that the loss of telomeres can be slowed down or sped up in the yeast S. cerevisiae. And importantly, they also show that some forms of environmental stress have no effect.
The authors looked at the effect of thirteen different environments on telomere length over 100-400 generations. They found that caffeine, high temperature and low levels of hydroxyurea lead to shortened telomeres, while alcohol and acetic acid lead to longer telomeres. It seems that for a long life, yeast should lay off the espresso and and try to avoid fevers, while enjoying those martinis and sauerbraten.
Romano and coworkers also found a number of conditions that had no effect on telomere length, with the most significant being oxidative stress. In contrast, previous studies in humans had suggested that the oxidative stress associated with emotional stress contributed to increased telomere loss; given these results, this may need to be looked at again. In any event, yeast can deal with the stresses of modern life with little or no impact on their telomere length.
The authors next set out to identify the genes that are impacted by these stressors. They focused on four different conditions—two that led to decreased telomere length, high temperature and caffeine, one that led to longer telomeres, ethanol, and one that had no effect, hydrogen peroxide. As a first step they identified key genes by comparing genome-wide transcript levels under each condition. They then went on to look at the effect of each stressor on strains deleted for each of the genes they identified.
Not surprisingly, the most important genes were those involved with the enzyme telomerase. This enzyme is responsible for adding to the telomeres at the ends of chromosomes. Without something like this, eukaryotes, with their linear chromosomes, would have disappeared long ago.
A key gene they identified was RIF1, encoding a negative regulator of telomerase. Deleting this gene led to decreased effects of ethanol and caffeine, suggesting that this gene is key to each stressor’s effects. The same was not true of high temperature—the strain deleted for RIF1 responded normally to high temperature. So high temperature works through a different mechanism.
Digging deeper into this pathway, Romano and coworkers found that Rap1p was the central player in ethanol’s ability to lengthen telomeres. This makes sense, as the ability of Rif1p to negatively regulate telomerase depends upon its interaction with Rap1p.
Caffeine, like ethanol, affected telomere length through Rif1p-Rap1p but with an opposite effect. As caffeine is known to be an inhibitor of phosphatydylinositol-3 kinase related kinases, the authors looked at whether known kinases in the telomerase pathway were involved in caffeine-dependent telomere shortening. They found that when they deleted both TEL1 and MEC1, caffeine no longer affected telomere length.
The authors were not so lucky in their attempts to tease out the mechanism of the ability of high temperature to shorten telomeres. They were not able to identify any single deletions that eliminated this effect of high temperature.
Whatever the mechanisms, the results presented in this study are important for a couple of different reasons. First off, they obviously teach us more about how telomere length is maintained. But this is more than a dry, academic finding.
Given that many of the 400 or so genes involved in maintaining telomere length are evolutionarily conserved, these results may also translate to humans too. This matters because telomere length is involved in a number of diseases and aging.
Studies like this may help us identify novel genes to target in diseases like cancer. And they may help us better understand how lifestyle choices can affect your telomeres and so your health. So if you have a cup of coffee, be sure to spike it with alcohol!
December 3, 2013
Our friend Saccharomyces cerevisiae has it pretty easy when it comes to sex. There is no club scene or online dating. Pretty much if an a and an α are close enough together, odds are that they will shmoo towards each other and fuse to create a diploid cell. No fuss, no muss.
Of course there aren’t any visual cues that indicate whether a yeast is a or α. Instead yeast relies on detecting gender-specific pheromones each cell puts out. The a yeast makes a pheromone and an α pheromone receptor, and the α yeast makes α pheromone and an a pheromone receptor. The way yeast finds a hottie is by looking for the yeast of the opposite sex that puts out the most pheromone.
This simple system is similar to ours in that gender is determined by gender specific gene expression. In humans this happens through the amounts of certain hormones that are made. For example, males make a lot of testosterone which turns on the androgen receptor (AR) which then turns a bunch of genes up or down. Both men and women have AR; men just make more testosterone, which causes it to be more active.
Yeast are simpler in that their mating loci encode transcription factors and cofactors that directly regulate a-specific and α-specific genes. Still, in both yeast and human, gender is determined by which genes are on and which are off.
Given how simple the yeast system is and how extensively it has been studied, you might think there is nothing else to learn about yeast mating. You’d be wrong. In a new study out in GENETICS, Huberman and Murray found that a gene with a previously unknown function, YLR040C, is involved in mating. They renamed this gene AFB1 (a-Factor Barrier) since it seems to interfere with a-factor secretion.
The way they found this gene was by creating, as they termed them, transvestite yeast that “pretended” to be the opposite mating type. One strain that they named the MATα-playing-a strain was α but produced a-specific mating proteins, while the other, the MATa-playing-α strain, was a but produced α-specific mating proteins. Sounds easy but it took a bit of genetic engineering to pull off.
The first steps in making the MATa-playing-α strain were to replace STE2 with STE3, MFA1 with MFα1, and MFA2 with MFα2. In addition, they had to delete BAR1 to keep it from chewing up any α factor that got made, and ASG7, which inhibits signaling from STE3. This strain still had the MATa locus, which meant that except for the manipulated genes, it still maintained an a-specific gene expression pattern.
Making the MATα-playing-a strain wasn’t much simpler. They had to replace STE3 with STE2, MFα1 with MFA1, and MFα2 with MFA2. In addition, they drove expression of BAR1 with the haploid specific FUS1 promoter and expression of the a-factor transporter STE6 with the MFα1 promoter. Maybe yeast isn’t so simple after all!
When Huberman and Murray mated the two transvestite strains to each other, they found that while these strains could produce diploid offspring, they weren’t very good at it. In fact, they were about 700-fold worse than true a and α strains! So what’s wrong?
To tease this out the researchers mated each transvestite to a wild type strain. They found that when they mated a wild type a strain to a MATa-playing-α strain, the transvestite’s mating efficiency was only down about three fold. By overexpressing α factor they quickly found that the transvestite strain’s major problem was that it simply didn’t make enough α pheromone. They hypothesized that perhaps differences in promoter strength or in the translation or processing of α-factor were to blame.
The reason for the low mating efficiency of the MATα-playing-a strain, however, wasn’t so simple. When Huberman and Murray mated the MATα-playing-a strain with an α cell, they found it was about 60-fold worse at mating. The first thing they looked for was how much a-factor this strain was producing. Because a-factor is difficult to assay biochemically, they used a novel bioassay instead and found that it secreted much less a-factor than did the wild type a strain. Further investigation showed that the transvestite strain produced something that blocked the ability of a-factor to be secreted.
By comparing the transcriptomes of MATa and MATα-playing-a cells they were able to identify YLR040C as their potential a-factor blocker. They went on to show that when this gene was present, a-factor secretion was indeed inhibited. They hypothesize that their newly named AFB1 may produce a protein that binds to and sequesters a-factor. It may be to a cells what BAR1 is to α cells, helping the yeast cell to sense the pheromone gradient and choose a mating partner.
When Huberman and Murray knocked AFB1 out of the MATα-playing-a strain, it now mated with a wild type α strain about five fold better than before. A nice increase, but it doesn’t completely correct the 60-fold reduction in this transvestite’s mating efficiency. Something else must be going on.
That something appears to be that the strain only arrests for a short time when it encounters α-factor. This would definitely impact mating efficiency, as it is very important that when a and α strains fuse they both be in the same part of the cell cycle. Pheromones usually stop the cell cycle in its tracks, but α-factor can’t seem to keep the MATα-playing-a cell arrested for very long. The researchers looked for genes involved in this transient arrest, but were not able to find any one gene that was responsible.
From all of this the authors conclude that there is a pheromone arms race raging in the yeast world. The most attractive yeast are those that make the most pheromone, so evolution favors higher and higher pheromone production. Just as people on the dating scene need to see past the makeup and trendy clothes to figure out who’s really the best partner, yeast need genes like BAR1 and AFB1 to parse out who is the best mate amid the ever increasing haze of pheromones.