New & Noteworthy

Retooling Yeast Factories

December 13, 2017


Imagine you own a rifle factory and you want to retool it to make solar panels instead. Or vice versa.

FactoryFloorNarrow

It would be silly to retool a factory one step at a time and yet that is what we do with microbial factories. (Wikimedia Commons)

It wouldn’t make a lot of sense to make small, incremental changes. Make one change, test to see how it works, make a second change, test to see how that works, and so on.

Not only is this inefficient and time consuming, but you may not end up with the best factory for making your new product. The best approach is to rip out the old machinery and replace it all at once. Or at the very least make many changes at a time.

The same sorts of problems exist for cellular factories—microbes we engineer to churn out useful products. Ideally we would want to get to an optimized, retooled cellular factory as soon as possible by making as many changes as we can at once. And yet, that is not how we currently do it.

Most of the time when we repurpose a microbial organism like Saccharomyces cerevisiae to make something new, we do it the slow way. We either do a screen to find genes that affect the process or change a known gene and analyze the results. We then use this mutant strain as a starting point to repeat the process. Over many cycles, we hopefully get to a yeast strain that can make something new or more efficiently.

Not only is this time consuming and inefficient, but we might miss some synergies that can happen when two or three genes are tweaked at once. This is no way to retool a factory!

In a new study out in Nature Communications, Lian and coworkers came up with a clever way to make many changes all at once in yeast. And they aren’t limited to just deleting genes either. They can transcriptionally activate, transcriptionally interfere, or delete specific genes. Which is where they got their clever acronym for their process—CRISPR-AID.

As you can tell from the name, their system uses the seemingly ubiquitous gene-editing tool CRISPR. And they use it in some very cool ways to get all three processes happening in a cell at the same time with different genes.

Conventional CRISPR works through a nuclease being guided by an aptly named guide RNA (gRNA) to a precise place in cell’s DNA through base pairing between the DNA and the gRNA.  Once there, the nuclease makes a double-stranded break and at least in yeast, the cell invariably repairs it using homologous directed repair (HDR). (Other eukaryotes tend to “fix” their DNA with the more error prone non-homologous end joining [NHEJ] pathway. Yes, yeast is superior in yet another way!)

Scientists have been able to tweak this system to activate or repress gene transcription by knocking out the nuclease activity of the CRISPR protein and adding back either an activation or repression domain. Now they have an RNA-guided, sequence specific transcription activator or repressor.

This is a great tool chest of gene editing tools but there is one problem—how to get the right protein to the right spot when all three use a gRNA for specificity. Like Mork from Ork turning to Mindy to help guide him on Earth, Lian and coworkers turned to a PAM to help guide these proteins to the right place in the yeast genome.

The protospacer adjacent motif (PAM) is a small part of the gRNA that binds the DNA and helps to pry it open so the rest of the gRNA can bind. Every gRNA has to have a PAM or the CRISPR protein won’t bind.

It turns out that different bacteria have different CRISPR systems that use different PAMs. The idea then is to use a CRISPR protein from different bacteria for each of the three different functions of the CRISPR-AID system.

The most common form of CRISPR protein is Cas9 from Streptococcus pyogenes (Sp), which has NGG for its PAM sequence. Lian and coworkers used a nuclease deficient form of this protein attached to a repressor domain to create their transcription interference tool dSpCas9-RD1152. (The d means the nuclease in inactivated, the Sp is the bacteria it came from, and the RD1152 is the repression domain.)

MorkMindyTradingCard

Like actress Pam Dawber’s character Mindy helping Mork find his proper place on Earth, so too does the PAM sequence guide the CRISPR protein to its proper place in the genome. (Joe Haupt)

For their deletion tool they turned to the Staphylococcus aureus (Sa) Cas9, SaCas9, which has NGRRT or NGRRN for its PAM sequence.

They turned to a different CRISPR protein, Cfp1 from Lachnospiraceae bacterium ND2006 (Lb), for their activation tool. This protein recognizes TTN as its PAM. They attached an activation domain to a nuclease-deficient form to generate dLbCfp1-VP.

All three proteins were stably integrated into a yeast cell. Now that their toolbox was full and their yeast cell ready, Lian and coworkers set out to show that these tools were effective.

They first used it on a known system, making beta-carotene in yeast. Previous work had shown that increasing the expression of the HMG1 gene, decreasing the activity of ERG9, and deleting ROX1 led to increased beta-carotene production.

They used CRISPR-AID to accomplish all three in one fell swoop. Not only did expression from HMG1 go up, expression from ERG9 go down, and ROX1 get deleted, but the resulting yeast also made more beta-carotene. They got a 2.8-fold increase in beta-carotene compared to only a 1.7 fold increase by hitting just one of the genes.

They next set out to increase expression of recombinant Trichoderma reesei endoglucanase II (EGII). Previous work had identified 14 genes that increased expression when activated, 17 that increased expression when inhibited and 5 that increased expression when deleted. Each of these was done as a single gene experiment.

Lian and coworkers created a library of gRNAs that could target each of the 1620 possible combinations to see if any combination would lead to a synergistic increase in activity. They found a combination that did just that—increased expression of PDI1, decreased expression of MNN9, and deletion of PMR1 led to the highest yield of EGII. The increased expression of EGII was beyond what any of the three did on their own.

So the system appears to be working well. Multiple targets can be upregulated, downregulated or deleted all at once.

While we are getting closer to being able to retool microbial factories as efficiently as brick and mortar ones, we aren’t there yet. That will require a deeper understanding of how a particular microbial organism works. And there is no better understood eukaryote than our old friend S. cerevisiae.

Like a good PAM sequence, Mindy helps Mork find his place in the world.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Genetics

Categories: Research Spotlight

Tags: gene editing, microbial factory, CRISPR

A SRTain Surprise in a Lipid Droplet

December 04, 2017


Barney Stinson from the sitcom “How I Met Your Mother” is right—the accidental curly fry in your plate of fries is a wonderful thing. It is unexpected but adds immensely to the French fry eating experience.

CleanerCurly

Just like this order of fries, spore wall lipid droplets have a pleasant surprise. (imgur)

In a new study in GENETICS, Hoffmann and coworkers show that spore wall lipid droplets, organelles used to move neutral lipids to the cell wall in the yeast Saccharomyces cerevisiae, have a surprise component, too. Instead of an errant curly fry in an order of fries, the greasy droplet has long chain polyisoprenoids made by Srt1p and its partner, Nus1p.

And in an unexpected twist, they are not used to glycosylate a protein as most polyprenols do. No, they upregulate the activity of a key protein involved in spore wall synthesis—Chs3p. Polyprenols may have more varied roles than scientists thought. A pleasant surprise indeed.

When a diploid yeast cell hits hard times, it undergoes sporulation to make sturdy spores to ride them out. And the spores’ sturdiness comes from the spore cell walls.

Yeast need many factors to make these resistant walls. One is Chs3p, the chitin synthase that makes the chitin that is used to make one of the four spore cell wall layers. Another is the lipid droplet that brings a variety of lipids and proteins to the site of the expanding cell wall.

Hoffmann and coworkers used transmission electron microscopy (TEM) to show that spores from strains deleted for SRT1 lack the chitosan layer of their cell walls. Additional experiments showed that this strain is missing this layer because it fails to make chitin as opposed to failing to be able to assemble chitin properly.

The Srt1p/Nus1p complex is a cisprenyltransferase—it adds 5-carbon isopentyl groups to farnesyl diphosphate to make polyprenols. In the next set of experiments, these researchers showed that an enzymatically inactive mutant of Srt1p, Srt1p-D75A, did not make the chitosan layer as well. This suggested that it is the long chain polyisoprenoids that Srt1p/Nus1p synthesizes that matters for spore wall formation.

In the final set of experiments, Hoffmann and coworkers showed that extracts lacking Srt1p had decreased chitin synthase activity. In other words, they showed that the activity of Chs3p is dependent on Srt1p.

Taken together, these results suggest that polyprenols like long chain polyisoprenoids may do more than simply facilitate the glycosylation of proteins. Here, they upregulate the activity of a protein involved in cell wall synthesis.

And we aren’t just talking yeast either. Most every living thing, including people, have polyprenol-derived dolichols which carry sugars to facilitate protein glycosylation. It may be that these types of molecules also have hidden, unexpected roles. Researchers may want to give these molecules a second look to see if they missed that extra curly fry.

Taking a man’s bonus curly is inexcusable.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Genetics

Categories: Research Spotlight

Tags: cell wall, spore, chitin, SRT1, chitosan

Squeezing Out a Tiny Bubble of DNA

November 20, 2017


Have you ever tried to walk back to the lab with too many plates between your thumb and middle finger? And had them be so compressed that they shoot out of your hands, spilling all over the floor? Yeah, me too.

PlatesInHand

In a process similar to this stack of plates collapsing when compressed, the basal transcription factor TFIIH causes small bubbles of opened DNA by compressing the DNA.

While this situation is bad for researchers, it turns out that the same sort of force may be helpful in cells by forcing the DNA open just a bit to get the ball rolling on transcribing a gene into messenger RNA (mRNA). The DNA is squeezed between two points so that a bubble of DNA pops open, leaving some single stranded DNA for RNA polymerase II (RNAP II) to get a hold of.

In a new study in Nature Structural & Molecular Biology, Tomko and coworkers show that Ssl2p, the double-stranded DNA translocase subunit of the basal transcription factor TFIIH, hydrolyzes dATP or ATP to force open around 5 or 6 base pairs of DNA. Once pried open, this small DNA bubble is expanded to 13 base pairs in the presence of NTP hydrolysis, presumably RNAP II beginning to transcribe the DNA. This stable open complex is now ready for promoter clearance and elongation.

These authors teased apart this mechanism using single-molecule magnetic-tweezers (henceforth referred to as tweezers). The idea is to stretch DNA out between two anchor points, add various components of the RNAP II machinery and various reagents, and to measure the changes in the stretched out DNA. Under the right conditions, these length changes directly reflect DNA unwinding.

Getting a gene read in a eukaryote like Saccharomyces cerevisiae is no easy task. A complicated mass of proteins called the preinitiation complex needs to form on the promoter first.

Tomko and coworkers loaded a subset of this complex, which included the TATA binding protein (TBP), TFIIB, TFIIF, TFIIH, and RNAP II, onto a piece of DNA with a single TATA box-containing followed by a strong initiation site. The authors arranged it so the 2.1 kilobases of DNA they used in their experiments was negatively supercoiled as is found in vivo.

In the first set of experiments, they added NTPs and dATP to their reactions. Most of the DNA did not show any changes as measured by the tweezers, but around 5% did.

There were two distinct populations of DNA in the subset that were active. Around 1/3 of the DNA showed  clear open and closed DNA transitions with the open DNA stretching over about 13 base pairs. The other 2/3 of the DNA showed longer-lived, smaller bubbles of around 5 or 6 base pairs of DNA. The authors interpreted the first as stable open complexes and/or elongation complexes and the second set as bubbles of DNA forced open by the preinitiation complex.

They repeated the experiment in the absence of NTPs and saw only the smaller bits of open DNA. And when they left out both NTPs and dATP, they saw no opened DNA at all.

So the small regions of opened DNA are dependent only on dATP hydrolysis making the Ssl2p subunit of TFIIH, a polypeptide known to hydrolyze dATP, the prime culprit for causing them. Which also makes sense, as Ssl2p is a DNA translocase that has been implicated in previous experiments in forcing the DNA to shorten between the preinitiation complex and downstream DNA. These experiments also showed that NTPs needed to be hydrolyzed to proceed to the next step, a stable open complex.

BubbleHeadCharm

Just as the bubble-head charm is needed to breathe underwater in the Harry Potter universe, so too are small DNA bubbles needed to get transcription started for RNAP II in ours. (Harry Potter Wiki)

Tomko and coworkers next wanted to determine if these smaller bits of open DNA play a meaningful role in transcription. They tackled this question with a couple of different experiments that both rely on the idea that opening the 5-6 base pairs of DNA is a necessary first step on the way to transcription.

In the first experiment, the authors saw that these smaller regions of open DNA had longer lifetimes at lower NTP concentrations. Remember, NTP hydrolysis is required to proceed to the next step, the 13 bp, more stable open complex. By decreasing the concentration of NTPs, the authors slowed this step down, causing the preceding step of the smaller 5-6 base pairs of opened DNA to hang around longer.

In the next set of experiments, they engineered DNA bubbles into the DNA that varied between 3 and 12 base pairs and tested what size bubble would allow transcription to proceed in the absence of TFIIH, the basal or general transcription factor they implicated in forcing these DNA bubbles open. Previous work had shown that TFIIH is unnecessary if the DNA has a 12 base pair region that is already opened.

They found that while 3 base pairs of opened DNA was insufficient to allow transcription in the absence of TFIIH (and dATP), an open region of 4 base pairs was. This is consistent with the idea that the 5-6 base pairs of opened DNA they saw in their experiments are relevant and that they are caused by the action of TFIIH.

Reading a gene is nontrivial in eukaryotes like us, and our friend yeast. The DNA is squeezed by TFIIH within the preinitiation complex to open just enough of the DNA for RNAP II to get a toe hold and get transcription rolling. A much better result than when a similar force gets your stack of plates rolling on the lab floor.

Don Ho probably isn’t singing about tiny, Ssl2p-mediated DNA bubbles but they too are fine.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Genetics

Categories: Research Spotlight

Tags: open complex, SSL2, RNA polymerase II, Preinitiation, TFIIH

Yeast Reveals a Vicious Circle Between Sugar and Cancer

November 10, 2017


The housing crisis that started in 2006 is a classic example of a vicious circle. As prices went down, some people couldn’t afford their mortgages and so were foreclosed upon. These foreclosed houses entered the market and caused prices to drop further which caused more foreclosures and so on. Only a Herculean effort by the government and the Federal Reserve was able to break this cycle.

ForeclosedHouse

A vicious circle between foreclosed homes and home prices popped the housing bubble starting in 2006. A similar vicious circle between glucose fermentation and cancer cell aggressiveness makes for particularly nasty cancers. (Wikimedia Commons)

Vicious circles can happen in biology too.

In a new study in Nature Communications, Peeters and coworkers have used our friend Saccharomyces cerevisiae to uncover one of these in cancer cells. And it turns out that this yeast is the perfect model organism for this study. (Funny how often that is true…)

Both S. cerevisiae and cancer tend to utilize their sugar through fermentation instead of respiration. This process in yeast is obviously great news for beer and wine drinkers. However, it is not such great news for people with cancer.

In the Warburg effect (named after the scientist who discovered it), the more aggressive a cancer is, the more sugar it ferments. This new study suggests that the consumption of sugar via fermentation and the aggressiveness of the cancer are related in a vicious circle.

Fermentation creates the byproduct fructose-1,6-bisphosphate (Fru1,6bisP) which, according to this study, activates the oncogene Ras which causes the cell to grow faster. As it grows faster, it ferments sugar faster creating more Fru1,6bisP which activates more Ras and so on. The cancer cells grow out of control until doctors apply some Herculean treatment to put a stop to it.

While yeast and cancer cells have a lot in common, wild type yeast isn’t quite up to cancer’s standards when it comes to cancer’s unbridled intake of glucose into the glycolysis pathway. These authors turned yeast a bit more cancer-like in this regard by deleting the TPS1 gene. Tps1p is a yeast hexokinase that tightly regulates yeast’s glucose intake into glycolysis.

Yeast cells without TPS1 deal poorly with glucose and need to be grown in galactose. When Peeters and coworkers added glucose to tps1Δ cells that had been previously grown in galactose, the cells activated Ras, a potent oncogene. This activation caused the cell to undergo apoptosis as assayed by cytochrome c release from the mitochondria, exposure of phosphadityl serine on the plasma membrane and generation of reactive oxygen species.

Deletion of hexokinase 2 eliminated this activation of Ras and suppressed the apoptosis. This suggests that perhaps some buildup of an intermediate metabolite in the glycolytic pathway might be to blame for the Ras activation.

The authors tested this hypothesis by removing the cell wall of wild type yeast cells and adding glycolysis metabolites to the resulting spheroplasts. They found that one metabolite, Fru1,6bisP, activated Ras. Supporting this finding, they found that deletion of both PFK1 and PFK2, two genes that encode phosphofructokinase 1, the enzyme responsible for making Fru1,6bisP, eliminated Ras activation in tps1Δ cells. Together, these data support the idea that Fru1,6bisP is the culprit behind Ras activation in yeast.

coffee_cycle

Vicious circles like the one behind the Warburg effect are everywhere! (Twisted Doodles)

All well and good, but what about human cancer cells? Is something similar going on there? You have probably guessed that the answer is yes.

The authors first depleted the level of Fru1,6bisP in two different cell lines by starving them of glucose for 48 hours. They then added back glucose, which has been shown to increase the levels of Fru1,6bisP in cells, and measured the activation level of Ras and two of its downstream targets, MEK and ERK. All three were transiently activated in both HEK293T and Hela Kyoto cell lines. Looks like Fru1,6bisP activates Ras in cancer cells as well.

So this might explain the Warburg effect that I mentioned earlier. The more glucose a cancer cell uses during fermentation, the more Fru1,6bisP it generates. And the more Fru1,6bisP that is made, the more Ras gets activated prompting the cell to grow faster. Which of course results in more glucose getting fermented and so on.

I don’t have time to go into the work these authors did to try to uncover the mechanism by which Fru1,6bisP activates Ras, but suffice it to say that it does not appear to be a direct interaction with Ras. Instead, it appears to work at least partially by disrupting the interaction between Ras and one of its guanine nucleotide exchange factors, Cdc25p (or Sos1 in humans).

And the story is not yet complete. There is still a lot of work to do in pinning down the specifics of this vicious circle. Yeast will undoubtedly be instrumental in helping us work through the rest of the details.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Genetics

Categories: Research Spotlight

Tags: cancer, Warburg Effect, PFK1, PFK2, TPS1, fermentation

Alliance of Genome Resources Opens Door with Version 1.0 Release

November 03, 2017


The Alliance of Genome Resources (the Alliance) announces the release of the Alliance of Genome Resources website 1.0 – providing unified access to comparative genetics and genomics data from the Alliance data resources (www.alliancegenome.org). The focus of the Alliance is to facilitate the use of these data towards better understanding of human biology and disease.

Alliance_Logo

The Gene Ontology Consortium and six Model Organism Databases have joined together to form the Alliance of Genome Resources.

The Alliance brings together the efforts of the major National Institutes of Health (NIH) National Human Genome Research Institute (NHGRI)-funded Model Organism Database (MOD) groups, and the Gene Ontology (GO) Consortium, in a synergistic integration of expertly-curated information about the functioning of cellular systems.

The MODs were created in the early days of the Human Genome Project in support of the major experimental models for human biology. The MODs currently included in the Alliance are the Saccharomyces Genome Database (SGD), FlyBase, WormBase, Mouse Genome Database (MGD), Rat Genome Database (RGD), and Zebrafish Information Network (ZFIN).  In addition, the Alliance includes the Gene Ontology (GO) Consortium. Now these groups will merge key activities and data representations, coordinating data retrieval and analysis, within a comparative perspective. Other MODs and related resources will be added to the Alliance going forward.

As part of this initial release, Alliance working groups have focused on the ability to easily access pages that summarize details of genes and diseases, with extensive representation of orthology data, and with access to multi-track JBrowse capabilities primarily for visualization of sequence data. Users recover gene details, functional information, and disease associations within a comparative perspective. As the integration of the MOD and GO teams progress with inclusion of additional data, the vision going forward includes the incorporation of other model organism information resources and other bioinformatic nodes within a common data platform, facilitating data recovery, analysis, and integration.

Categories: News and Views

Tags: Alliance of Genome Resources

Something from Nothing: A New, Structured Protein from Noncoding DNA

November 01, 2017


In the Rick and Morty episode Mortynight Run, a gaseous life form pulls off what alchemists had been trying to do for centuries. It literally creates gold out of thin air!

GaseousGold

Like this gaseous being converting oxygen into gold, Saccharomyces cerevisiae has converted noncoding DNA into a protein with at least some secondary structure. (Reddit)

Our ever faithful friend, Saccharomyces cerevisiae can’t do that (at least not yet). But what it has done is create a protein, Bsc4p, out of noncoding intergenic DNA. And while not gold, it is a fully functioning protein.

In a new study, Bungard and coworkers show that this recently evolved gene, BSC4, encodes a protein that can fold into at least a partially defined structure. This matters as there was some debate about whether newly generated proteins could attain a defined structure or if they would remain as intrinsically disordered proteins (IDPs). A reasonable debate, given how rare structure is among amino acid sequences and how plentiful IDPs are in a cell.

Bsc4p is a great protein to study in this regard as there is very strong evidence that it has evolved relatively recently in S. cerevisiae, but not in other closely related species. And it definitely does work in a S. cerevisiae cell. While not essential, some genetic studies (including this one) indicate that it plays an important role in DNA damage repair pathways.

Here is an example from S. paradoxus of the noncoding sequence that the S .cerevisiae BSC4 gene almost certainly sprang from:

gtg TCT GTA ATT CTA CGG AAA AGT AAA CAA AAA AAC TGT AAT TGC ATA ACG AGC AAT TTA TAT ACA ATA CAC ATA GAA AGA CTT TCG CTC tga TGT CCG AAC TGC CAT TGT CAT TGG AGA AAA TCC TTA TGT GGA GTG GAG TTC CCT GCA GGT TAT TTT CAG AGA AAA CGT GGT TAC AAA AAG GGA CCA GAT TCG CCC tag CTT ACA ACT CGC TTG AAT CAT CTT TAT GCC AGA CCT TTC AAC GCC GCG ACC CCA AAA ACA taa ATG CTG AGT CAC CAT GGT GCT GGG CGC TGT CGC TGT CGC GCT GTT CCT TTC CGA GAA AAG CAC GGC AAC AAC AAC AAC AGT CCA TAT GAC CAA AAA AAA AAT AAC CGC AAA TGG CAG tga AAT GCA ATT ATC ATT GTA TAC GA?

In order for this sequence to become a gene coding a protein, at a minimum the first lowercase, dark orange codon needs to be mutated to ATG, a start codon, and the rest of the lowercase, dark orange codons need to be mutated away from being stop codons.

This seems to be part of what happened in S. cerevisiae:

ATG TCT ATT GTG CTA CGG AAG AGT AAC AAA AAA AAC AAA AAC TGC ATA ACA AGC AAG TTT TAT ACA ATA CAC ATT ATA AAA ATT TCT ACT CCG GTG TTC CGA GCT CCC ATT GCC ATT GGA GAA AGC CCT TAT GTG GAG TGG AGC TGC CTA CAG GTT GTT TTC AGG AAA GAC ATG GTT ACA AAA AAG ACG ACA TTC GCC CAA CTT ATC ACT CGC TTG AAC CAC TTT TTA TGC CAA GCC CTT AAA CGC CGC GAC TCA AAA ACA TAC ATA CTG TGC CGC ACG GCA GTT TTT GGC GCT ATG ACA CCC TTT TCC CCA AGA AAA TCG CAT ATT AAC AAC AAA TTA CCC ATG CAA CCC AGG AAA AAA AAA ATA GTC ATT ATA TAC GTA GTG CGC TTT CAT TGA

Through a few small changes, we now have a 131 amino acid polypeptide where before we had some noncoding DNA between LYP1 and ALP1.

Bungard and coworkers use a variety of techniques to show that this newly evolved protein has structure. Not as much structure as many proteins that have been around longer, but more than many of those IDPs.

PhilStone

It is as if Saccharomyces cerevisiae has found the philosopher’s stone but instead of changing lead to gold, it turns noncoding DNA into a gene that codes for a partially structured protein. (Wikimedia Commons)

Consistent with protein structure, Bsc4p forms compact oligomers under native conditions that are partially resistant to proteolysis, has a far UV circular dichroism (CD) spectra consistent with beta sheets, has a buried tryptophan, Trp47, that becomes solvent accessible under denaturing conditions, as measured by tryptophan fluorescence, and has a near UV spectra consistent with a hydrophobic core. However, they found no evidence of any significant interactions between the secondary structures to form a single three dimensional shape, in the protein.  In other words, no evidence of a tertiary structure.

And that wasn’t the only sign that Bsc4p wasn’t a mature, fully structured protein. For example, that near UV CD that showed a hydrophobic core, was weak in intensity, which is consistent with at least “partially molten character.” And Bsc4p bound certain dyes: Congo red, Thioflavin T, and ANS, in a way consistent with some molten globule and/or amyloid character.

So we have a bit of a mixed bag with Bsc4p. One way to think about it is as a young protein still developing its ultimate three dimensional structure. Or, it could be that for the job it does, this is all the structure it needs.

In any event, it is definitely a newly evolved protein with at least some structure which shows that this can indeed happen. Sometimes Mother Nature can make structured proteins from noncoding DNA. Like that gaseous being on Rick and Morty, producing gold out of thin air.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Department of Genetics

Categories: Research Spotlight

Tags: noncoding DNA, protein structure, BSC4, newly evolved gene

Transcription Factors: Kings of the Genome Jungle

October 20, 2017


Imagine that Jane is in trouble on the edge of the jungle. She needs to be saved soon or she will be sent back to Europe (which she does not want).

june71

Tarzan got to Jane in time by swinging on vines. Just like Mig1p needs to “swing” on DNA to get to its binding sites in the yeast genome on time. (DisneyClips)

Tarzan knows he can’t get there in time by running along the jungle floor. But of course he has a trick up his sleeve—swinging from vine to vine!

He gets there in time, fends off her would-be kidnappers, and saves the day. All because he transferred from one strand of vine to another instead of “sliding” along the jungle floor.

A new study by Wollman and coworkers shows that at least two transcription factors in Saccharomyces cerevisiae, Mig1p and Msn2p, seem to share a lot in common with Tarzan. In a process termed intersegment transfer, they get to where they need to in the genome by “swinging” from DNA segment to DNA segment instead of just sliding along the DNA.

And transcription factors like these need to get to the right place in time to save the cell from outside threats. Just like Tarzan had to swing from vine to vine to save Jane.

This sort of approach would not necessarily work well with just a single transcription factor with a single DNA binding domain. It would sort of be like a one-armed Tarzan—it is hard to take advantage of swinging from vine to vine without at least two arms!

The authors argue that Mig1p gains its “extra arms” through joining together into a cluster. Now each Mig1p can bind DNA and drag the other transcription factors with them. A neat solution to the one arm problem.

Their basic approach is to use fluorescence microscopy to follow a GFP-Mig1p fusion protein in a single cell, something that has only become possible recently. What they found was that there were two populations of transcription factors—a diffuse set of smaller molecules and distinct, larger clusters made up of multiple Mig1p’s.

The clusters appeared to be the ones doing the work in the nucleus. Kinetic studies showed that they stayed in one place in the nucleus for over 100 seconds which is consistent with the clusters and not the monomers being bound to the DNA.

OK so this transcription factor tends to clump up into clusters and it looks like these clusters are the ones regulating gene expression. They also showed that a second transcription factor, Msn2p, did the same thing.

The authors next set out to see if this approach made sense for genetic regulation by running simulations of Mig1p finding its sites in the nucleus as either monomers or as clusters. It made sense to form clusters.

A lot of previous work has been done in S. cerevisiae in terms of the three dimensional map of the genome and where Mig1p DNA binding sites were located in this mesh of DNA. And in the course of their studies, Wollman and coworkers were able to estimate how many Mig1p molecules were in yeast cells and how many were in clusters.

MultiArmedMonster

This mythical beast is more like a cluster of Mig1p proteins with its multiple arms representing multiple DNA binding domains with which to grab strands of DNA. (Wikimedia Commons)

They now had all the information they needed to run their simulations. When they crunched the numbers, they found that clusters fit their data much better than monomers (R2 = 0.75 vs. R2 < 0).

The final step was to work out what part of Mig1p was involved in forming the clusters. To do this, the authors compared Mig1p and Msn2p, the second transcription factor they studied that also formed into clusters, and looked for structural regions they might have in common.

What they found was both proteins had a highly disordered region. For Mig1p, it was at the C-terminus and for Msn2p it was at the N-terminus.

The hypothesis is that these disordered regions, which are both at the opposite end of the protein from the DNA binding domain, interact and form ordered structures that enable clusters to form.  Wollman and coworkers used circular dichroism to show that when Mig1p was put in conditions that favor cluster formation, there was a transition consistent with unstructured protein becoming structured.

What we seem to have is a cluster of transcription factors connected in the middle with their DNA binding domains pointing out. This rolling cluster can more easily hop from DNA strand to DNA strand to find the right spots to bind.

Without this mechanism, Mig1p couldn’t get to where it needs to in time. It is as if Tarzan had 6-9 arms circling his body so he could get to Jane even more quickly.

With the wide range of tools available and our deep understanding of how yeast works and how a yeast cell is organized, our trusted ally S. cerevisiae again teaches us something fundamental about how our biology works.

Mig1p may be able to swing like Tarzan but it can’t yell like him. Or like Carol Burnett!

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Department of Genetics

Categories: Research Spotlight

Tags: free diffusion, MIG1, intersegment transfer, Saccharomyces cerevisiae, transcription factor

Divide and Prosper

October 09, 2017


A Swiss Army knife is a marvelous thing. It is a bunch of useful tools all wrapped up in something small enough to put in your pocket.

army-2186_1920

Like a Swiss Army knife, sometimes you can’t tweak one function of a multifunctional protein without disrupting its other functions. (Pixabay)

Need to cut a string? It has the scissors to do that. How about cutting that little piece of wood? Yup, the tiny saw can do that. It has lots of tools that can do lots of things.

Now of course, it isn’t perfect. You would not want to use those scissors for cutting fabric or to use that saw to cut a 2 by 4. And there really isn’t any easy way to make a better pair of scissors or a better saw in the Swiss Army knife without wrecking the other tools in it.

If you make a better pair of scissors, you’ll have a worse saw. And vice versa.

That is why if you want to improve the saw, you need to isolate it away from the other tools. This holds true for the scissors and yes, even the knife.

Multifunctional proteins can have similar constraints. Tweak one function and it can mess up the other functions. A better saw makes a worse pair of scissors.

Nature doesn’t usually improve on a protein’s function by pulling out the part of the protein that does a specific job and making it better. No, instead, she duplicates the gene with the instructions for the protein. Now Nature can optimize each function of each gene copy—both improved functions can still happen albeit with separate proteins.

It is like having two Swiss Army knives and improving the saw on one and scissors on the other. You still have scissors even when your improvements on your saw have wrecked the attached scissors. They are just on the other knife.

In a new study in GENETICS, Hanner and Rusche explored this idea of “subfunctionalization” using the Saccharomyces cerevisiae proteins Sir3p and Orc1p. The genes that encode these proteins are thought to have arisen from a single ancestral gene after a whole-genome duplication around 100 million years ago.

The ORC1 gene is highly conserved in eukaryotes and serves an important role in DNA replication. In addition to this role, in S. cerevisiae, it also works to silence the mating type loci by recruiting the SIR complex to either HMRa or HMRα. Sir3p is also involved in silencing genes including HMRa or HMRα, but it does its job by recognizing deacetylated histones.

The basic strategy of this study is to compare Sir3p from S. cerevisiae with KlOrc1p, an Orc1p homolog from the yeast Kluyveromyces lactis that is involved in both DNA replication and gene silencing.

Since K. lactis did not undergo the whole genome duplication that S. cerevisiae did, the idea is that KlOrc1p represents the multifunctional ancestral protein (or at least what it has evolved into over the past 100 million years or so). Sir3p is the saw while KlOrc1p is the saw as part of the Swiss Army knife.

A MATa strain of S. cerevisiae lacking Sir3p cannot mate because its HMRα locus is not repressed. And a MATα strain of S. cerevisiae lacking Sir3p cannot mate because its HMRa locus is not repressed.

In the first experiment, Hanner and Rusche wanted to see if Orc1p from yeasts that did not undergo the whole genome duplication could rescue strains of S. cerevisiae lacking Sir3p. In other words, do these non-duplicated proteins have Sir3p functions or did Sir3p evolve new functions once freed from its DNA replication constraints?

These authors showed that KlOrc1p could not rescue mating in a strain of S. cerevisiae lacking Sir3p (but still containing its own Orc1p). They took this even further and showed that Orc1p from two other non-duplicated yeasts, Lachancea kluyverii and Zygosaccharomyces rouxii, could not rescue the S. cerevisiae strain either. Sir3p in S. cerevisiae has functions that none of these “ancestral” Orc1p’s have.

musical_saw_2

Not easy to play a song like this with a saw from a Swiss Army knife. (Wikimedia Commons)

The next step was to create chimeras between Sir3p and KlOrc1p to identify the parts of Sir3p that developed new functions.  Both proteins consist of four broad subdomains: AAA+, winged helix, N-terminal bromo-adjacent homology (BAH), and a rapidly evolving linker.

The researchers swapped domains between the two proteins and found that a Sir3p with an AAA+ subdomain from the KlOrc1p did not allow a S. cerevisiae strain lacking Sir3p to successfully mate. So Sir3p has gained some new function in AAA+ that the ancestral protein lacked. The authors were able to narrow down this region a bit more—the AAA+ subdomain lacking its first two alpha helices worked too.

Hanner and Rusche used chromatin immunoprecipitation assays to show that this chimeric protein did not make it to the HMRa or HMRα loci or to other repressed genes. They also looked at mRNA levels and showed that genes that should have been repressed weren’t. This Sir3p with the AAA+ subdomain of KlOrc1p clearly cannot do the job of Sir3p.

And it isn’t because of the loss ATPase function of the AAA+ domain in Sir3p. While Sir3p has an ATPase domain, it does not have any discernible ATPase activity. When Hanner and Rusche replaced the KlOrc1p ATPase domain with the ATPase domain of Sir3p, this protein still couldn’t rescue the S. cerevisiae strain deleted for SIR3. The ATPase activity of KlOrc1p does not explain its inability to rescue the strain.

It is known that the AAA+ domain of Sir3p interacts with Sir4p. Another possible explanation for the failure of KlOrc1p to complement the loss of Sir3p may be that KlOrc1p has over the last 100 million or so years lost the ability to interact with the S. cerevisiae version of Sir4p. This does not seem to be the reason as adding in the K. lactis version of Sir4p, KlSir4p, to the S. cerevisiae strain deleted for Sir3p, did not allow the strain to mate. Inability to interact with Sir4p does not seem to explain the failure of KlOrc1p to rescue the strain.

The authors hypothesize that once freed of the constraints of the multifunctional Orc1p, the AAA+ region of Sir3p gained the ability to interact with deacetylated histones. This will need to be tested in future studies.

There you have it. Once freed of having to focus on two things, Sir3p was able to evolve new functions like recognizing deacetylated histones. Like a saw being freed from a Swiss Army knife so that it can sing.

Don’t try this with the saw from a Swiss Army knife.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Department of Genetics

Categories: Research Spotlight

Tags: SIR3, ORC1, subfunctionalization, gene duplication

Sensing Disturbances in the Membrane

September 25, 2017


In the original Star Wars movie, Darth Vader knows Obi-Wan is aboard the Death Star because he feels a “disturbance in the force.” The highly conserved protein Ire1p does something similar with disturbances in the membrane of the endoplasmic reticulum (ER) in Saccharomyces cerevisiae.

Comic Con reveals Spangdahlem's true identity

He can sense disturbances in the force. Ire1p can sense disturbances in the ER membrane. https://media.defense.gov

Ire1p isn’t sensing the presence of a Jedi master of course. Instead, it is sensing when the membrane is under stress through either the buildup of unfolded proteins in the ER lumen or membrane and/or through changes in the membrane itself.

And unlike Darth Vader, who upon sensing the presence of Obi-Wan seeks him out and destroys his physical form with a light saber, Ire1p instead forms dimers and larger oligomers when it senses a disturbance. This activates the unfolded protein response (UPR), which results in the activation of a bunch of genes so that the cell can start to deal with its unfolded proteins and perturbed membrane mess. (And no, unfortunately, the cell does not do this with tiny light sabers!)

Previous work had established the part of Ire1p that responds to unfolded proteins. But scientists had yet to nail down what part of the protein senses membrane problems. Until now that is.

In a new study in Molecular Cell, Halbleib and coworkers have not only worked out what part of Ire1p senses disturbances in the ER membrane in our old friend S. cerevisiae, but they may also have figured out how the protein does it.

At first blush, it would seem to make sense that the single, highly conserved transmembrane helix (TMH) of Ire1p would be important for sensing membrane issues. After all, it is in intimate contact with the membrane!

Previous mutation studies showed that the TMH alone probably isn’t the key player. Halbleib and coworkers confirmed this result by replacing all thirteen amino acids of the TMH in Ire1p with leucines and showing that this mutated protein was still able to respond to membrane problems caused by either too much dithiothreitol (DTT) or too little inositol.

A closer look at the protein revealed a conserved region of Ire1p that overlaps with the TMH and is predicted to form an amphipathic helix (AH). CD spectroscopy showed that an oligopeptide of identical sequence to this region formed an alpha helix in micelles but not in aqueous solution—it needs a membrane environment to become an alpha helix.

These authors used mutational analysis to show that the hydrophobic side of this AH is the part of the Ire1p involved in sensing problems in the ER membrane. For example, two specific mutations on the hydrophobic side, F531R and V535R, negatively affected a yeast cell’s ability to respond to disturbances in the ER membrane. Yeast containing Ire1p with a mutation on the hydrophilic side, R537E, however, had no such problem and responded fine. Ire1p function was also compromised when the researchers extended their leucine replacement of the transmembrane domain by three more residues into the AH.

So Halbleib and coworkers have seemingly identified the key part of Ire1p that can sense changes in the membrane. Now, they set out to examine how this part of the protein actually does the sensing.

ForceChoke

I find your lack of oligomerization disturbing. http://starwars.wikia.com/wiki/Force_choke

Right away the authors could see that the mutants that disrupted the ability of Ire1p to sense membrane disturbances also did not oligomerize under membrane-perturbing conditions in a living yeast cell. This makes sense as oligomerization is required to marshal the cellular forces needed to deal with these ER problems.

They confirmed that oligomerization could be affected by membrane composition with some in vitro experiments. They first made what they called a minimal sensor construct—the amphipathic helix and the transmembrane helix attached to the maltose binding protein. They reconstituted this sensor into eight different liposomes with varying lipid compositions.

Using continuous-wave paramagnetic resonance spectroscopy they were able to show that the ability of their minimal sensor to oligomerize depended upon the membrane composition. The more similar the composition was to a stressed membrane, the more likely their sensor was to oligomerize. This is exactly what we would expect if the AH was the key player in sensing membrane problems.

So certain mutations and membrane situations can affect the oligomerization state. But this still doesn’t really get at why Ire1 proteins oligomerize in perturbed membranes.

To try to figure this out, they turned to molecular dynamic simulations.

They found that the AH tended to compress the lipid bilayer and that the more stressed the membrane, the more energy this compression costs. They also found that if two Ire1p molecules happened to bump into each other and dimerize, that the energy costs went down.

In other words, stressed membranes tend to force Ire1p molecules to oligomerize to reduce the energy cost of being in the membrane. And of course oligomerization triggers the unfolded protein response (UPR)!

Ire1p senses disturbances in the membrane because the energy cost of remaining a monomer in stressed membranes is too high. Much more plausible than the midi-chlorians that supposedly let Darth Vader sense disturbances in the force!

The unfolded protein response (UPR) is involved in many diseases including cancer and diabetes. 

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Genetics

Categories: Research Spotlight

Tags: UPR, membrane sensing, IRE1

Cheerio Containment

September 11, 2017


In the store the other day I saw a kid in a stroller covered in Cheerios, an opened, near empty Ziploc-type plastic bag next to him. It looked like one of those cheap, knock off brands that are much harder to close properly.

p_101384655

Just like a poorly closed zipper means more spilled Cheerios, a looser H1 clamp means more transcripts. Image from http://www.parents.com.

Other than keeping the Cheerios out of reach, his parents had two options here. The first was to use two bags, putting the cheaper re-sealable bag into another one. The odds are much better that at least one of them would be closed properly.

The second option was to get a re-sealable bag with a better zipper, like an actual Ziploc. With a better baggie, the parents had a much better chance of making sure that the bag was sealed.

In a new study out in GENETICS, Lawrence and coworkers show that yeast cells go for the second approach when it comes to the histone H1 and gene expression regulation. Rather than regulating gene expression by controlling the number of molecules of H1 in a cell, the yeast instead regulates transcription by controlling how well H1 can bind DNA. And it looks like the strength of binding is influenced by the acetylation of H1’s more famous histone cousins—the histones found in the nucleosome.

Almost everyone has heard of the histones that make up the nucleosome, that structure of DNA wrapped around a core of eight histone proteins. H1 is a bit less famous. It binds the linker DNA, or the DNA between some of the nucleosomes.

When bound, H1 tends to repress nearby genes. It acts like a clamp that keeps the nucleosome in place.

This gene repression is probably why our friend Saccharomyces cerevisiae has only one H1 per 4-37 nucleosomes as compared to the one H1 per nucleosome in vertebrates. After all, S. cerevisiae’s genome has a much higher gene density. If it had one H1 for every nucleosome it might not have any genes turned on!

You may have noticed that scientists don’t have a good handle on an exact number of H1 molecules in a yeast cell. One of the first things these authors did was a careful study to try to nail down just what that number is. They came up with 18.9 +/- 1 nucleosomes/H1.

The next step was to figure out how a yeast cell controls H1 to regulate gene expression. There are two general ideas about how a cell might go about this.

In the first, the cell controls the amount of H1 it makes. The more H1, the more genes get repressed.

In the second, H1 is controlled by the acetylation of the core histone proteins. The acetylation loosens the core histone’s grip on the DNA making it harder for H1 to bind.

It would be very hard to do experiments in most beasts to try to tell which model is correct. For example, mammals have 11 genes for H1 proteins. This makes changing the overall amount of H1 nontrivial to say the least.

This is where the ultimate model organism, S. cerevisiae, can come in and save the day yet again (#APOYG). Since this yeast has only one H1 gene, HHO1, it is much simpler to tweak the amount of H1 protein in a cell.

Lawrence and coworkers put HHO1 under the control of the GAL promoter and found that in the presence of galactose, the extra H1 had a severe effect on growth. These yeast were unhappy with around three times as much H1 protein, even though there was not much of an effect on chromosomal structure as measured by micrococcal nuclease assays.

cheerios_in_ziploc

Low acetylation on this Ziploc. Image from General Mills.

The researchers next looked for genes which when deleted, rescued the growth defect of yeast making too much H1. They did this by transforming their GAL-HHO1 plasmid into the collection of ~4700 nonessential gene knockout strains.

Histone deacetylases (HDACs) were the most interesting class of genes they found that could rescue the growth defect. By deleting an HDAC gene, histone acetylation increases (because the deacetylation decreases), implying that H1 can be regulated by the acetylation status of histones. The loss of the HDAC increased the amount of acetylation, which made it harder for H1 to bind, and so repress transcription.

They confirmed this finding in a couple of ways with the most interesting one using a couple of different mutants of histone H3. One mutant had its acetylation sites in the H3 tail changed to glutamines, which mimics a histone in a perpetual state of acetylation. This mimic of an acetylated H3 tail partially rescued the growth defect of too much H1.

So it looks like the effect of H1 can be regulated by the acetylation status of its more famous histone kin. More acetylation results in a looser grip, which makes for more transcription. Like a cheap plastic baggie with a zipper with a looser grip making Cheerios more likely to spill, H1’s looser grip due to more acetylation results in more transcription and more transcripts being let loose into the cell.

by Barry Starr, Ph.D., Director of Outreach Activities, Stanford Genetics

Categories: Research Spotlight

Tags: HHO1, nucleosome, H1, histones, gene regulation

Next