Saturday, June 23, 2007

Borderline mutated

Although the difference between CLL with mutated and unmutated immunoglobulin variable region heavy chain (IgVH) genes is well established and this distinction is recognised as one of the most important prognostic variables, the choice of 98% as the limit of the unmutated subset was arbitrary, based on possible polymorphisms that might produce that degree of variation.

What is a polymorphism? Everybody is different. The difference between you and me is the sum of the polymorphisms in all our genes. The classic polymorphism is the difference between normal haemoglobin and sickle haemoglobin. A single amino acid switch from glutamic acid to valine at position 6 in the beta chain of haemoglobin is enough to cause a devastating disease. But every protein in the body is subject to such polymorphisms, though most of them are completely silent. Some will determine whether you are blood group A or O, or whether you have blue or brown eyes, or how long your ear lobes are. The immunoglobulin genes are no exception, but it presents us with a problem. How do we know whether a variation from the accepted sequence in a particular IgVH gene is a somatic mutation (as part of germinal center induced variation to improve antibody-fit) or a simple polymorphism. The obvious answer is to sequence the IgVH genes in a different tissue that isn’t subject to somatic hypermutation such as the neutrophils or skin fibroblasts. But that is a time consuming business and as a short cut scientists decided that it probably didn’t matter very much and decided that up to 2% sequence variation might be due to polymorphisms, so patients with more than 98% sequence homology would be regarded as unmutated.

Not everybody agreed on the 98%. A German group chose 97% and a British group 95% as a more appropriate cut-off. However, when David Oscier from our lab in Bournemouth decided to investigate this by looking at IgVH gene homology in leukaemic cells and granulocytes from the same patient, it became apparent that even small numbers of mutations were caused by somatic hypermutation rather than polymorphisms.

Individuals whose IgVH genes have only a few somatic mutations comprise a small proportion of CLL patients and until recently it has not been possible to study enough patients in this group to distinguish a different prognostic impact of choosing the threshold for unmutated status at >97% or >98% homology. A final difficulty is the discovery that the use of the VH3-21 gene is associated with a poor prognosis whether of not there are somatic mutations. This is especially problematic because such cases frequently have between 96 and 99% sequence homology.

In order to resolve these difficulties we have made a retrospective survey of 310 patients with CLL who have passed through our hands in the past 30 years and who have had their IgVH genes sequenced. We have compared outcomes of patients with different degrees of sequence homology. There were only four whose tumor used the VH3-21 gene, none of which fell in the disputed area of 97-98% homology; two had > 99% and two <97% sequence homology. Patients were observed until treatment was indicated because of symptoms or evidence of progression, and then treated according to best practice of the day; largely with chlorambucil prior to 1990 and increasingly with purine analogues or combinations including purine analogues since then.

There were 99 patients with 100% sequence homology, 22 with between 98 and 99.9%, 22 with between 97 and 97.9%, 24 with between 96 and 96.9%, and 143 with <96% homology with germ line genes. There have been 139 deaths of which 79 were determined to be unrelated to CLL. Survival curves are shown in figure 1. The median survivals, censored for unrelated deaths, were 102 months for patients with 100% IgVH gene homology; 132 months for those with 98-99.9%, 184 months for those with 97-97.9% and not yet reached for those with 96% or <96%. The difference between 100% and 97-97.9% was statistically significant (p=0.002) as was the difference between 97-97.9% and <97% (p<0.0001). However, the differences between 100% and 98-99.9% and between 98-99.9% and 97-97.9% did not reach statistical significance. The survival curves for those with 96-96.9% homology and <96% were virtually identical.

Using treatment-free survival as an alternative end-point gave very similar results. Median times to treatment were 35 months for 100% homology, 36 months for 98-99.9%, 156 months for 97-97.9%, and 272 months for <97%. The difference between 100% and 97-97.9% was statistically significant (p=0.001) as was the difference between 97-97.9% and <97% (p=0.0003), but the differences between 100% and 98-99.9%, between 98-99.9% and 97-97.9%, and between 96-96.9% and <96% were not.

The pattern revealed by this study is not a continuous gradation with survival increasing with increases in the number of mutations. Those with 97%-97.9% homology comprise a mixture of benign and malignant cases rather than a homogeneous group with moderate malignancy.

It is not understood how accumulations of somatic mutations affects prognosis in CLL. The initial explanation that in those with mutated IgVH genes the cells had passed through the germinal centre while the cells of those with unmutated IgVH genes had not seems unlikely to be true. Tumour cells from both types of patients most closely resemble memory B cells; both express CD27 and both are now thought of as antigen-experienced.

An alternative possibility is that the mutator mechanism is impaired in CLLs with unmutated IgVH genes. The anomalously high expression of activation-induced cytosine deaminase (AID), an enzyme necessary for somatic hypermutation and Ig class switching, in cases with unmutated IgVH genes may be implicated. It has been suggested that high levels of AID may result in loss of substrate specificity and the development of mutations in c-MYC, PAX-5 and RhoH genes which are associated with more aggressive forms of the disease. It would be interesting to see whether the level of AID correlates with the degree of somatic hypermutation.


Anonymous said...

Thanks for a great post! I must admit I've never thought too greatly about the dividing line, just accepting as fact the cut-off. (Sadly I'm now more concerned with treatments rather than the science behind CLL.)

I have followed the antigen-experienced/unexperienced debate, and concur with the mainstream thinking. The fact that CLL cells are similar to memory B cells has always struck me as significant.

The thinking that there may be a mutational malfunction present prior to the onset of CLL seems to be something that could be tested. Is this the case in the healthy population there is some variation in the ability to produce mutations? Perhaps (and easier and less costly) it would be possible to test mutated/unmutated CLL patients to see if there is a difference in the variety of mutations?

Could this explain at least part of the defective immune function of the CLL patient? I've wondered if the immune defect exists BEFORE the development of CLL. Perhaps there is such a defect.

This study answers a lot of questions. Thanks for discussing it.

Anonymous said...

By the way; is this data available from, say, UC San Diego? I am unmutated, but don't know the percentage. Is this data routinely available?

BTW, I found out that they don't test for lipoprotein lipase, which has been suggested to be an independent prognostic indicator.

Terry Hamblin said...

If you know you are mutated then the lab that did the test will know how mutated.

Lipoprotein lipase has to be measured by a molecular technique that is beyond the scope of most laboratories, and this has deterred most labs for undertaking it.

Tina said...

That's great information about borderline mutations ever seen.