Thursday, December 5, 2013

Diffuse Large B Cell Lymphoma (knowing your ABC's and GCB's)

Growing up, I religiously read the comic strip, “The Far Side.”  In one strip I remember clearly, there was a woman reading her veterinary medicine textbook.  In chapter 9 she came to equine medicine where the remedy for just about anything that bothered a horse was exactly the same.  The captions said, “Like most students, Doreen breezed through equine medicine.”  I’ve posted a link here for a quick laughSometimes treating patients with diffuse large b cell lymphoma (DLBCL) can feel the same  - R-CHOP for all (here is a description of R-CHOP and my personal approach to DLBCL).  
If you look at the most recent update of the world health organization classification of blood cancers, they identify thirteen different types of “aggressive lymphoma.”  Most patients are familiar with the most common subtype lovingly called DLBCL-NOS (not otherwise specified).  The “plain vanilla” DLBCL proves to be far and away the most common version of the disease.

Despite being lumped together as a single entity, if you pop the hood and look deeper inside DLBCL-NOS, there is a lot of biologic heterogeneity – those differences are starting to look more and more important.  A number of years ago, one of my colleagues (who I still insist is the smartest guy I’ve ever met – Ash Alizadeh at Stanford) took advantage of a brand new technology called “Gene Expression Profiling” to look at a bunch of DLBCL samples and got his findings published in the journal Nature.  Other labs (Lou Staudt at NIH and Margaret Shipp at Dana Farber) had similar findings – it is pretty robust.  This allowed scientists to take cancer samples and test them to see which genes were turned on and turned off (a term we call “expression”) and evaluate tens of thousands of genes at a time.

What they all found was that there were two “main” subtypes of DLBCL-NOS.  Those two subtypes were the “Germinal Center DLBCL” and the “Activated B Cell DLBCL” (GCB and ABC subtypes respectively).  Purists who know the science well could fault me for not being completely precise here, but that will do for now.  When normal B cells undergo proliferation in response to discovering the “germ” they were destined for all their life, they do so in a specialized region of the lymph node called the “germinal center.”  GCB-DLBCL-NOS appears to have similar genes turned on and turned off (ie expressed) whereas ABC-DLBCL-NOS looks like a cell that has left the germinal center in an “activated” state.”

Why does this matter?  It turns out that these biologic differences have significant clinical impact.  If you segregate ABC from GCB DLBCL, the ABC do quite a bit worse than the GCB in terms of overall survival and response to R-CHOP chemotherapy. 

Since there is such a big clinical impact, you would think we would all know if our DLBCL patients were ABC or GCB subtype – but we don’t!  It turns out that gene expression profiling (GEP) is pretty expensive and it has to be done on biopsies handled a certain way that most surgeons / pathologists don’t do.  Instead of GEP testing, pathologists figured out how to make the determination of ABC vs. GCB using tests that are a lot cheaper and easier to use – called “immuno-histo-chemistry” (or IHC).  Any pathologist who looks at cancer specimens knows how to do IHC.  You essentially take a small slice of the tumor / lymph node / marrow / etc., stain it with an antibody of interest, and use a marker on the antibody to help you know if it stuck after you washed the heck out of it.  ABC has different IHC staining than GCB DLBCL when you look at markers like CD10, BCL-6, MUM-1, and so forth – so just about any lab can do this testing – but they don’t always do it!  There are two main reasons why many don’t.  First, the test isn’t nearly as reliable as GEP testing.  In fact, there are several different antibody combinations that can help make the IHC distinction, but get a bunch of pathologists together and they will only agree about 70% of the time when they do the stains themselves.  The second reason is when I come back to the far side comic strip.  At least for now, you pretty much have the same R-CHOP no matter what the test shows – but that is starting to change.

For patients interested in a purely “prognostic” test, I suggest getting this one done.  It distills all of the gene expression profiling down to two genes and can be ordered easily today.

Since fewer than half of ABC DLBCL patients are likely to be cured with their R-CHOP chemotherapy, lots of pharmaceutical companies are interested in building a better version of R-CHOP.  Quite a few drugs have been explored in DLBCL in the last few years that seem to have preferential benefit in the ABC subtype of DLBCL.  Ibrutinib yields a 40% response rate as a single agent in relapsed ABC-DLBCL but only a 5% response rate in GCB subtype.  For Celgene’s lenalidomide, it is more like 50% versus 5%.  Millennium’s bortezomib can be added to R-CHOP and when you look at the patients with ABC vs GCB, they do almost identically suggesting that the drug overcomes the negative impact of the ABC subtype of the disease.  Indeed, adding ibrutinib to R-CHOP gives a 100% response rate in preliminary studies, and adding lenalidomide to R-CHOP makes the ABC look ever so slightly better than the GCB subtype.  Not surprisingly there is intense interest in ongoing phase III studies that restrict enrollment to patients with ABC subtype DLBCL to see if any of these three drugs can be added to standard R-CHOP.

My prediction?  I bet several of these end up crossing the finish line and getting approved by the FDA (disclaimer: I am not currently involved in ANY of these studies and have not seen any preliminary data).  So what happens if two or three such drugs actually get approved – how would a doc and patient choose which combo to take?  Now we go back deeper into the biology.

If we look at gene expression profiling or use immunohistochemistry staining to categorize DLBCL-NOS into ABC or GCB we are really only using surrogates for a more fundamental process – what mutations have occurred at the DNA level.  Several publications have recently come out where the entire genome of DLBCL has been “sequenced” and when you look at the data, you realize there are a handful of mutations that seem to recur.  Those mutations often determine which genes are expressed, and how the cell looks under the microscope when you use IHC.  In essence GEP and IHC are merely proxies for the underlying mutations.

Here is where it gets really interesting and makes a lot of sense.  The ABC subtype of DLBCL appears to have “chronic active signaling” through the B-Cell receptor (BCR).  This has some similarities to what we have seen in CLL through different mechanisms.  If you look at the BCR signaling pathway (think of an electrical circuit), you can create a “short circuit” at any number of steps along the way, yet the result seems to be a light that either won’t turn on or off appropriately.  In this case, that is a protein called NF-kB which regulates the expression of a bunch of important B cell genes.

If that short circuit is really high up in the pathway (CD79 stuck in the “on” position) or low in the pathway (CARD-11 mutation) you get the same activation of NF-kB, but turning it off at different places may make a difference.  BTK is a protein that lies just “downstream” of CD79 in the BCR signaling pathway and lies “upstream” of CARD-11.  It is also the target for ibrutinib.  You can have ABC-DLBCL where there is a mutation in EITHER CD79 or CARD11.  Interestingly, very preliminary work makes it look like ibrutinib works well when there is a CD79 mutation but it isn't clear yet how well it works when  there is a CARD-11 mutation.  This makes a lot of sense.  To get to NF-kB, BCR signaling from abnormal CD79 has to go through BTK but BCR signaling that starts from abnormal CARD-11 does not.  That would perfectly explain why ibrutinib might work in one case but not the other (if larger numbers of patients make that theory hold up).

So my challenge to my pharma colleagues is the following: figure out which mutations confer activity for your drug (as some are trying to do right now) so that the docs know what to do once your drug is approved.  Molecular diagnostic tests are now available that enable you to test for each of these mutations all at once.  This should be in their interest anyway because the FDA loves "companion diagnostics" for selecting individuals for therapy.  I predict that in the future, DLBCL management will go like this:
Patient walks in with new diagnosis of DLBCL

Since therapy is often needed very quickly, the patient will get their first cycle of R-CHOP

While the three weeks between cycles has passes, the patient will have their cancer sent for analysis and they will find out what unique mutations they have

The patient will come back for the second dose of therapy and an appropriate targeted drug will be added based upon the mutation analysis

In some cases no “pathway specific” mutation will be found.  In these cases, there are other drugs working through the system that may be added like the antibody drug conjugates (see here and here).  Pharma companies developing such drugs would be wise to know where their drug works even if it is felt not to be mutation specific so that they can pick up the pieces for the patients who don’t get ibrutinib, lenalidomide, bortezomib or others.

R-CHOP has served us well for quite a few years.  We are spoiled in NHL management that we get to cure a lot of our patients.  It is really fun as a doc to have a patient walk in doing great with their cancer a fading memory.  Unfortunately “a lot”doesn’t feel that great if you are not one of the ones who is cured.  I think we are getting really close to making some important steps forward in the management of DLBCL-NOS and it will be driven by molecular information that leads to targeted therapies that are specific to the patient.

Thanks for reading!