Casting a beady eye on kinases

A huge area of drug discovery is targeting protein kinases. There are about 500 protein kinase-like proteins in the human proteome. Some of these are probably not active (pseudokinases; review [paid]) and periodically there are claims of protein kinase activity in novel proteins, but that's the ballpark. An increasing number of drugs target these, with Gleevec as perhaps the best known but a large parade of others coming forward.

Most kinase-targeting drugs compete with ATP in the active site. The ATP-binding site of kinases shows a lot of conservation, and so cross-reactivity is a big topic in the field. What is desired for specificity depends on the target & disease & one's tolerance for risk. Gleevec was originally touted as being laser-focused on BCR-ABL, but it actually hits a number of kinases and many of these have yielded new markets, such as KIT for gastrointestinal stromal tumors. Being 'dirty' may be useful in oncology, where many kinases may be contributing to the tumor's growth & survival. On the other hand, in chronic diseases one probably wants a really focused drug (or at least can't tolerate one that isn't).

I got involved a little in kinase screening back at MLNM. The workhorse in the industry are in vitro assays using purified kinases. These are useful and can be run en masse, but everyone knows in vitro isn't always predictive of in vivo. Furthermore, despite diligent efforts by a number of vendors, not every kinase is available. So, the field is ripe for development of new approaches, especially ones which explore the compound in vivo.

In an ideal world, there would be a complete panel of biomarkers specific for each kinase which could be used to measure the impact of a compound on every kinase-regulated pathway in the cell. That's a long, long ways off -- only a few kinases really have good, reliable assays & many are essentially uncharacterized.

The latest Nature Biotech has a nice paper from CellZome on a proteomic approach to the problem & an accompanying News & Views item from a top mass spec person (either link prior requires Nat Biotech subscription, but Cellzome has the paper for free also).

The strategy is to derivatize beads with promiscuous kinase-binding compounds and use these to pull down bound kinases for identification by Mass Spec. Such has been described previously, particularly by the company which Cellzome acquired which had been doing this work. Big advances in this paper is to use iTRAQ labeling reagents to enable accurate quantification of the bound proteins and to use a competitive-binding format to assay test compounds.

The competitive binding angle is clever. Past efforts tried to derivatize the compound of interest to link it to the beads. This has many undesirable properties: the linkage may change the binding properties, each compound must be worked out separately. etc. The new work identified a set of standard promiscuous-binders which can be used to get a large fraction of the kinase repertoire. This same set can be used with just about any compound. By using these in a competitive format, where what is measured is how much the compound of interest disturbs the binding profile of the kinobeads, quantitative measurements can be made. Entire binding curves can be pulled out from a few experiments -- and binding curves for every kinase reliably pulled down by the beads. Slick!

Their coverage of the kinase world is quite good, though not perfect. In a set of experiments they pulled down 307 kinases. While that isn't everything, it's a lot -- and some of the rest may be pseudokinases or not expressed in any of the tissues they looked at. However, some probably just aren't bound well by the reference compound set. Whole small subtrees are missing from their Figure 2 -- examples of the missing include all the GPCR kinases (BARK, GRKs, etc), the two TLKS1, 2/4 polos, the WNKS, none of the 3 Akts, a number of miscellaneous cell cycle kinases (CDC7, BubR1&R2, etc. Some of those are mighty interesting, but of course the solution is to find more compounds to put on the beads.

What's also nice is that the assay isn't limited to kinases -- lots of other stuff comes down too. Initially it was apparently clogged with heat shock proteins (using a n ATP analog as the probe). But, in the current format they found a good sampling of non-kinases, and for Gleevec identified a candidate off-target with some known biology.

So what's the catch? Well, there are a few caveats. First, it is a binding assay, so hits need to be followed up to demonstrate actual inhibition. Second, it is going to be ATP-site specific. That covers most current kinase inhibitors, but there are probably many out there that work outside the ATP site, and this method will be blind to that (or to off-targets bound not by ATP-mimicry). Third, many of things being pulled down won't have much known about them -- do you worry about it or not?

As described in the M&M, the assay requires 5 mL of cell lysate -- not tiny, but not whopping. So this probably wouldn't be applied to every compound coming out of medicinal chemistry, but perhaps to either characterize representative members of particular lead families and to characterize compounds that are quite a ways down the med chem path.

One other interesting bit at the end: instead of adding the test compounds to lysates, they added the compounds & waited a few hours. As they note, many kinase inhibitors have slow off-rates -- they stick to their target rather fiercely. So, the assay can be run after a delay. The peptide mixtures could then be subjected to phosphopeptide enrichment & identification, enabling the phosphorylation state of downstream kinases to be probed & examined in relation to the concentration of compound used.

New Breast Cancer Molecular Diagnostic

The Cancer Genetics blog has a post on the approval of Veridex's new RT-PCR test for breast cancer spread.

What was emphasized in the Globe article which is striking is that this test can potentially be performed while the patient is still on the operating table, avoiding a delay between screening test & initiating follow-up testing. If this holds true, then this is an example of molecular diagnostics really having a big impact in a major health problem. As with any diagnostic test, the key question is specificity & sensitivity aka false positives and false negatives. The key study had 300ish patients in it, which is just a small puddle compared to the ocean of breast cancer patients.

Veridex, which is owned by J&J, has some other cool technologies cooking, including some to sift tiny numbers of cancer cells from the bloodstream, cells which have escaped from the primary tumor or metastases. Since getting clinical samples can be a serious challenge, this technology is pretty amazing.

Imaging gene expression

In one of my first posts I commented on the challenge of obtaining samples for microarray and other biomarker work. Getting samples for microarrays is at best difficult, painful to the patient and only a little dangerous to them; in many cases the samples are simply unobtainable. Getting a broad range of samples from multiple sites, or a time series is going to be very rarely feasible.

With this backdrop, a recent paper in Nature Biotechnology is quite stunning. Indeed, it is a bit of a surprise that it didn't show up in the mother ship or Science: the paper is well written, audacious in design and shows very nice results.

Using actual liver cancer patients the paper correlates contrast-enhanced CT (aka CAT) imaging features to gene expression patterns detected by microarrays using samples from the same patients. While these patients had to go through biopsies, the approach holds out the hope of calibrating imaging assays for future use.

The imaging-microarray connections have many intriguing possibilities. Some of the linked microarray patterns have clear therapeutic associations, such as cell cycle genes and VEGF. Such an imaging approach might, with much further validation, enable appropriate selection of therapeutic agents -- such as Avastin to target VEGF.

The paper also notes the challenges that lie ahead. The choice of liver cancer was no accident: liver tumors tend to be large and well-vascularized, making them straightforward to image using CT. Some of the imaging features found are generic to tumors, but others have some degree of liver specificity. Expression program to image feature mappings may vary from tumor to tumor.

One potential side-effect of this study would be to increase biopharma interest in liver cancer. Liver cancer is a scourge outside of the Western world (perhaps driven by food-borne toxins) but is not in the top of deadly cancers in the U.S. According to some 2002 figures from the American Cancer Society, liver cancer in the U.S. is about 17K new cases and about 15K fatalities -- a horrible toll, but far less than 160K annual lung cancer deaths. One big attraction for companies is potential payoff, but another is the potential for accelerated development decisions. Being able to subset patients based matching drug mechanism to biology inferred from imaging is potentially a powerful means to do that.

More lupus news

Hot on my previous rant around lupus is some more news. Human Genome Sciences has announced positive results for its Lymphostat B drug in lupus. I won't go into detail on their results, other than to comment that the study size is large (>300), the trial is a Phase II double-blind placebo controlled trial (open label, single arm trials are much more common for Phase II -- HGS isn't taking the easy route) but these results haven't yet been subject to full peer review in a journal article.

Lymphostat B has a number of unusual historical notes attached to it. It is in that very rarified society of discoveries from genomics which have made it far into therapeutic clinical trials -- there are other examples (not on hand, but trust me on this!), but not many. It doesn't hurt that it was in a protein family (TNF ligands) which suggested a bit of the biology (e.g. a cognate receptor) & has led to quite a bit of biology which is in the right neighborhood for a lupus therapy (B-cell biology); most genomics finds were Churchillian enigmas.

Second, this is a drug that initially failed similar trials -- but HGS conducted a post-hoc subset analysis on the previous trial. However, instead of begging their way forward (such analyses get all the respect due used cat litter, but that doesn't stop desperate companies from trying to argue for advancement) they designed a new trial using a biomarker to subset the population. If their strategy works, it is likely that doctors will only prescribe it to this restricted population. HGS has, in effect, decided it is better to treat some percent of a small population than risk getting approval for 0% of a larger one -- a bit of math the pharmaceutical industry has frequently naysayed.

HGS and their partner GSK still have a long way to go on Lymphostat B. Good luck to them -- everyone in this business needs it, especially the patients.

Small results, big press release

The medical world is full of horrible diseases which need tackling, but you can't track them all. For me, it is natural to focus a touch more on those to which I have a personal connection.

Lupus is one such disease, as I have a friend with it. Lupus is an autoimmune disease in which the body produces antibodies targeting various normal cellular proteins. The result can be brutal biological chaos.

The pharmaceutical armamentarium for lupus isn't very good. Anti-lupus therapies fall into two general categories: anti-inflammatory agents and low doses of cancer chemotherapeutics (primarily anti-metabolite therapies such as methotrexate). Few of these have been adequately tested in lupus, and certainly not well tested in combination. The docs are flying by the seat of their pants. The side effects of the drugs are quite severe, so much so that lupus therapy can be an endless back-and-forth between minimizing disease damage & therapy side effects.

One reason lupus hasn't received a lot of attention from the pharmaceutical industry is that we really don't understand the disease. It is almost certainly a 'complex disease', meaning there are multiple genetic pathways that lead to or influence the disease. Different patients manifest the disease in different ways. For many patients, the most dangerous aspect is an autoimmune assault on the kidneys. but for my friend the most vicious flare-ups are pericarditis, an inflammation of the sac around the heart. These differences could reflect very different disease mechanisms; we really don't know.

We need to understand the mechanisms of lupus, so it is with interest I read items such as this one: New biomarkers for lupus found. The item starts promisingly

A Wake Forest University School of Medicine team believes it has found biomarkers for lupus that also may play a role in causing the disease.

The biomarkers are micro-ribonucleic acids (micro-RNAs), said Nilamadhab Mishra, M.D. He and colleagues reported at the American College of Rheumatology meeting in Washington that they had found profound differences in the expression of micro-RNAs...

So far, so good -- except now things go south

...between five lupus patients and six healthy control patients who did not have lupus.

Five patients? Six controls? These are exquisitely tiny samples, particularly when looking at microRNAs, of which there are >100 known for human. With so few samples, the risk of a chance association is high. And are these good comparisons? Were the samples well matched for age, concurrent & previous therapies, gender, etc?

Farther down is even more worrisome verbiage

In the new study, the researchers found 40 microRNAs in which the difference in expression between the lupus patients and the controls was more than 1.5 times, and focused on five micro-RNAs where the lupus patients had more than three times the amount of the microRNAs as healthy controls, and one, called miR 95 where the lupus patients had just one third of the gene expression of the microRNA of the controls.

Fold-change cutoffs are popular in expression studies, because they are intuitive, but are generally meaningless. Depending on how tight the assays are, fold changes of 3X can be meaningless (in an assay with high technical variance) and ones smaller than 1.5X can be quite significant (in an assay with very tight technical variance). Well-designed microarray studies are far more likely to use proper statistical tests, such as T-tests.

And one last statement to complain about

The team reported the lesser amount of miR 95 "results in aberrant gene expression in lupus patients."

Is this simply correlation between miR 95 and other gene expression -- which suffers both from the fact that correlation is not causation and that with such small samples gene expression differences will be found from pure chance. Are these genes which have previously been shown to be targets of miR 95? Has it been shown that actually interfering with miR 95 expression in the patient samples reverts the gene expression changes?

Of course, it is patently unfair for me to beat up on a scientific poster of preliminary results for which I have only seen a press release - one hopes that before this data gets to press a much more detailed workup is performed (please, please let me review this paper!). But, it is also patently unfair to yank the chains of patients with understudied diseases with press releases that take a nub of a preliminary result and headline it into a major advance.

Phosphopallooza.

Protein phosphorylation is a hot topic in signal transduction research. Kinases can add phosphate groups to serines, threonines & tyrosines (and very rarely histidines), and phosphatases can take them off. These phosphorylations can shift the shape of the protein directly, or create (or destroy) binding sites for other proteins. Such bindings can in turn cause the assembly/disassembly of protein complexes, trigger the transport of a protein to another part of the cell, or lead to the protein being destroyed (or prevent such) by the proteasome. This is hardly a comprehensive list of what can happen.

Furthermore, a large (by some estimates 1/4 to 1/5) amount of the pharmaceutical industries efforts, including those at my (soon to be ex-) employer Millennium, are targeting protein kinases. If you wish to drug kinases, you really want to know what the downstream biology is and that starts with what does your kinase phosphorylate, when does it do it, and what events do those phosphorylations trigger.

A large number of methods have been published for finding phosphorylation sites on proteins, but by far the most productive have been mass spectrometric ones (MS for short). Using various sample workup strategies, cleverer-and-cleverer instrument designs, and better software, the MS folks keep pushing the envelope in an impressive manner.

The latest Cell has the latest leap forward: a paper describing 6,600 phosphorylation sites (on 2,244 proteins). To put this in perspective, the total number of previously published human phosphorylation sites (by my count) was around 12,000 -- this paper has found 50% as many as were previously known! Some prior papers (such as these two examples) had found close to 2,000 sites.

Now some of this depth came from many MS runs -- but that in itself illustrates how this task is getting simpler; otherwise so many runs wouldn't be practical. The multiple runs also were used to gather more data: looking at phosphorylation changes (quantitatively!) over a timecourse.

One this this study wasn't designed to do is clearly assign the sites to kinases. Bioinformatic methods can be used to make guesses, but without some really painful work you can't really make a strong case. And if the site shouldn't look like any pattern for a known kinase -- good luck! There really aren't great methods for solving this (not to say there aren't some really clever tries).

Also interesting in this study is the low degree of overlap with previous studies. While the reference set they used is probably quite a bit lower than the 12K estimate I give, it is still quite large -- and most sites in the new paper weren't found in the older ones. There are in excess of 20 million Ser/Thr/Tyr in the proteome and many are probably not phosphorylated, but certainly a reasonable estimate would be north of 20K are.

For drug discovery, the sort of timecourse data in this paper is another proof-of-concept of the idea of discovering biomarkers for your kinase using high-throughput MS approaches (another case can be found in another paper). By pushing for so many sites, the number of candidates goes up substantially, since many sites found aren't modulated in an interesting way, at least in terms of pursuing a biomarker. This is noted in Figure 3 -- for the same protein, the temporal dynamics of phosphorylation at different sites can be quite different.

However, it remains to be seen how far into the process these MS approaches can be pushed. Most likely, the sites of interest will need to probed with immunologic assays, as previously discussed.

In vivo nanobodies.

The new print issue of Nature Methods showed up, and it is rare for this journal not to have a cool technology or two in it. If you are in the life sciences, you can generally get a free subscription to this journal.

Antibodies are cool things, but also complex molecular structures. They are huge proteins composed of 2 heavy chains & 2 light chains, all held together by disulfide linkages. Expressing recombinant antibodies is not a common feat -- it is very hard to do so given the precise ratio & the folding required. Trying to express them inside the cytoplasm would be even trickier, as the redox potential won't let those disulfides form.

Camels & their kin, however, have very funky antibodies -- only a single heavy chain. I've never come across the history of how these were found -- presumably some immunologist sampling all mammals to look for wierd antibodies. Because of this structure, they are much smaller & don't require disulfide linkages. In fact, the constant parts of the camelid antibody can be lopped off as well, leaving the very small variable region, termed a nanobody.

The new paper (subscription required for full text, alas) describes fusing nanobodies to fluorescent proteins & then expressing them in vivo. Since only a single chain is needed, the nanobody coding region can be PCRed out & fused to your favorite fluorescent protein. The paper shows that when expressed in cells, these hybrids glow just where you would expect them to. The ultimate vital stain for any protein or modification! With multiple fluorescent protein of different colors, multiplexing is even theoretically possible (though not approached in this paper).

Of course, one is going to need to generate all those nanobodies. There is already a company planning to commercialize therapeutic nanobodies (ablynx). Perhaps another company will specialize in research tool nanobodies -- ideally without the nanoprofits and nanoshareprices which are all too common in biotechnology!

Nanowesterns: The future of signal transduction research?

Western blots are a workhorse of biology. When everything goes right, they allow for interrogating the state & quantity of a protein in a cellular system. They can be exquisitely sensitive and specific; Western blot assays have long been used as the definitive test for a number of medical conditions, particularly HIV infection. Given the right antibody, you can detect anything, including miniscule amounts of phosphorylated proteins. And, to a first approximation, they are quantitative.

A Western blot involves several steps. First, the samples of interest are placed in a denaturing buffer, causing the proteins to unfold & disaggregate. The unfolding is performed by large quantities of detergent (primarily SDS, which also shows up in your toothpaste, laundry detergent, dishwashing liquid, etc -- the stuff is ubiquitous) and the disaggregation is assisted by some sort of sulfhydryl compound to destroy disulfide bridges. Such compounds are uniformly smelly, except to a lucky few (I had a graduate school colleague who was smell-blind for them).

Now the samples are loaded on an SDS-PAGE gel, which uses electricity to separate the proteins by size -- approximately. In theory. Once they are separated, the proteins are transferred to a membrane by osmosis or electrophoresis perpendicular to the first direction. The extra protein binding sites on the membrane are then blocked, often with Carnation non-fat dry milk (I kid you not; the stuff is cheap & works). An antibody for the target of interest is added, and then an antibody to detect the first antibody; this one carries a label of some sort. Occasionally it is a third antibody which detects the 2nd (which bound the first) -- each level can enhance sensitivity. The appropriate detection chemistry is run & voila! You have a Western blot. Between all the steps after the blocking are lots of washes to remove excess reagents.

The beauty of a Western is that the technology is pretty cheap & simple -- I did a bunch of Western's in my senior thesis in a lab that ran on a shoestring budget -- and I'm all thumbs in the lab. The truly amazing part is that Westerns today are run pretty much the same way. You might buy pre-poured gels, but the basics are all the same.

The problems are legion.

First, this is a decidedly low-throughput assay scheme -- typical gels have maybe two dozen lanes for running. This is one reason it is used as a confirmatory test for HIV and other infections; large scale testing is out of the question.

Second, it is very labor intensive. Setting up the transfer from gel to membrane is inherently a manual process, but somewhat surprisingly it still seems uncommon to automate the later steps or even the washing. During a short lived Western blot process improvement project I initiated, I discovered that the folks running the blots both disliked the washing but also found it a social activity -- everyone is doing something mindless, so there is time to talk (simple fly pushing in a Drosophila lab is similar; the lab I was in almost always had NPR going in the background).

Third, they can require a lot of tuning. Different extraction ("lysis") buffers for the initial extraction, different gel or running conditions, different membranes, antibody dilutions, etc. -- these are all variables one can play with on the blot. Some rules of thumb are out there based on the location of the protein or how greasy it is, but it is clearly more art & lore than science. Some proteins never seem to work. Many result in big messes -- which is another advantage of Westerns, as you have the electrophoretic separation perhaps parsing the mixture into uninteresting bands & the one you want -- which may well be much fainter than the junk. And, of course, the gels don't always run the right way. Too hot -- trouble. Not poured evenly -- trouble. And please don't drop them on the floor!

Some of the trouble comes from the antibodies, but that is easily a topic for another time. But most is inherent in the Western scheme (no, there was no Dr. Western -- but there was a Dr. Southern and the other compass blots are plays on that).

But they are still extremely useful. Some folks have tried to push the envelope within the boundaries of a conventional Western. Perhaps the best example of this has been commercialized at Kinexus, which has pushed multiplex Westerns to amazing limits. They work carefully to identify sets of antibodies which will not interfere with each other & which also generate non-overlapping bands. One way to think about this is a really good Western antibody generates a single band in the same spot on the gel -- which means the rest of the blot is wasted data. Kinexus tries to maximize the amount of data from one gel. But this is a lot of work.

A new publication (free!) describes an approach that has a bit in common with the Western, but in many ways is altogether a different beast. The work is done by a startup called Cell Biosciences.

The slab gel is replaced by capillaries -- easy to control on the thermal side. SDS-PAGE is replaced by isoelectric focusing (IEF). Instead of blotting to a membrane, the separated proteins are locked onto the wall of the capillary. But other than everything being different, it's a Western!

Isoelectric focusing is a technique for electrophoretic separation of proteins. Instead of size, which SDS-PAGE sorts on, IEF uses protein charge. Each protein contains some amino acids which have positive charge and some with negative charge, plus postively & negatively charged ends. Post-translational modifications can further stir the pot: phosphorylation adds two negative charges per phosphate, and something big like ubiquitin tacks on a complex mess. On the other end, some modifications, such as acetylation, may replace a charged group with an uncharged one. In a gel with a pH gradient & subject to an electric field, the proteins will migrate to the pH where they have no charge -- the positively ionized and negatively ionized groups are in perfect balance.

An advantage of this pointed out in the nanowestern paper is that you can load a lot more sample on a gel. For a size separation, only a narrow band of sample can enter the gel because the separation is based on different sized proteins traveling at different speeds. Because IEF is an equilibrium method, you can actually fill the entire capillary with sample and then apply the electric field. This has important sensitivity implications.

The paper also describes a whole apparatus for automating the whole shebang; quite a contrast from an ordinary western. The model described runs only a dozen capillaries, but modern DNA sequencers routinely run hundreds simultaneously so there is plenty of room to grow. Each capillary detects one analyte for one sample, so with hundreds you could process hundreds of samples or detect hundreds of analytes, or some interesting middle ground.

Capillaries are also intriguing because they are at the heart of many lab-on-a-chip schemes. This paper might suggest the notion of a multi-analyte integrated chip.

The paper also describes using multiple fluorescent peptides as internal standards. These are synthesized in the opposite handedness as natural peptides, this doesn't change their IEF properties, but does make them unpalatable to proteases that might be present in the sample (though in general you use cocktails of inhibitors to prevent those proteases from attacking your sample).

The authors describe using a single antibody to assess the phosphorylation states of two related proteins, ERK1 and ERK2. Such determinations can be challenging on a Western if the two run closely with each other. In an SDS-PAGE gel, the behavior of phosphorylated proteins is maddening -- sometimes they run with the unphosphorylated form and sometimes they form new bands (a "phosphoshift"). Murphy's law rules; whichever behavior you don't want is the one you get! With IEF, phosphorylation should always give a strong phosphoshift, as a weakly ionizable side chain (hydroxyl on a Ser, Thr or Tyr) is replaced with a strongly acidic phosphate.

The paper describes using their system with a few proteins. That's a good start, but I suspect most people will want to see more. A lot more. And with other modifications. Ubiquitination would be particularly interesting, both because it is a big modification and because ubiquitin chains are often form. Presumably this will lead to a laddering effect. Also interesting to look at are more complicated phosphorylation systems than the ones examined here, with tens of phosphorylation sites rather than a handful. A reasonable guess is that the approach will still count sites, but if you want to distinguish them, which generally you will, you will still need specific antibodies for each site.

One last plus of their scheme, which they place right in the title & is the source of the nano moniker. The system is very sensitive, at least with the one analyte tested. This is important for rare or small samples (more on samples in another post). They even claim they might be able to push it from 25 cells down to 1 cell. If this is true, or even if 25 cells is achieved consistently with many antibodies, this will be an impressive feat & make this technique very attractive for signal transduction research.