Herceptin—controversy with clever science
Herceptin is a drug that has attracted a fair measure of attention and controversy in recent years. It is used in the treatment of a certain type of breast cancer where some deem it to be of considerable benefit. Unfortunately, it is also very expensive and has adverse effects on the hearts of some patients. While we will consider some of these points later, the drug itself merits attention. Far from your everyday pill, it is termed a “humanised monoclonal antibody” and is one of the first such compounds to enter clinical use. Monoclonal antibodies were a Nobel Prize-winning invention of the late 1970s. To appreciate their ingenuity you need to understand some immunology.
Antibodies are proteins, naturally produced by cells (termed B cells) in the vertebrate immune system, that bind to large foreign molecules. Foreign molecules? Yes, the immune system has the ability to discriminate between every large molecule (macromolecule) that is a normal component of the body (self molecules), and foreign macromolecules, which are usually part of a bacterium, virus or fungus infecting it. The immune system attacks foreign macromolecules.
The targets of the attack are usually proteins, but they can be other large organic molecules, such as nucleic acids (DNA) and polysaccharides (complex sugars). Molecules that provoke an immune response are termed antigens. Small molecules are not good antigens in that they don’t usually provoke immune attacks. That said, the piece of a foreign antigen to which an antibody binds is quite small—perhaps a dozen amino acids out of a string of several hundred—and is termed an antigenic determinant (see diagrams opposite).
A key idea is that the antibody is recognising and binding to a particular three-dimensional shape. Usually the antibody binding is very, very specific for the shape (or antigenic determinant) it recognises. The part of the antibody that binds the antigenic determinant is called the complementarity determining region (CDR) because it is thought to have a complementary shape—a bit like a lock and key—and these parts of an antibody are also extraordinarily variable. One macromolecule may have a lot of different antigenic determinants to which different antibodies may bind (see diagram).
Proteins on the cells of transplanted organs and on transfused blood cells are also seen as foreign by the immune system and antibody production is part of the process of tissue rejection.
The body will also make antibodies against any foreign protein a researcher may inject. Antibodies, because of the very specific nature of their binding, have become essential tools in biology for detecting, identifying and quantifying other biomolecules. Laboratory animals worldwide—usually rabbits and mice—are regularly immunised and subsequently have blood samples taken to procure antisera against proteins of interest to science and medicine. Antisera against snake venoms are well known.
For instance, a cancer researcher may inject membranes isolated from human breast cancer cells into a mouse and the mouse will make antibodies against the human proteins. Some of those antibodies can then be obtained from the mouse in a blood sample and used, for instance, to track the distribution of membrane proteins in normal and cancerous tissues. Antibodies can be given coloured or radioactive tags that make them detectable.
But there is more we need to understand about antibody production before we can appreciate monoclonals.
In some ways the immune system works pretty much like a blind gunner in a tower, firing off bullets from his machine gun constantly in all directions, in the hope that he will hit an invading terrorist he can’t even see. New B cells—millions an hour are constantly being produced from bone marrow and (this is important), each makes only one sort of antibody that binds one random shape the size of a single antigenic determinant. Counterintuitive though it seems for something as specific and effective as an antibody response, B cells make the CDR regions of antibodies using completely random selections from a broad menu of gene fragments which has nothing at all to do with the shapes of incoming proteins. The immune system’s philosophy is, let’s make antibodies to every possible shape and some of them are bound to work!
Each B cell carries many copies of the one type of antibody it manufactures on its surface. If it collides with an antigen that bears a shape complementary to that of its surface antibody, they bind. This key interaction triggers that B cell to start dividing, and then to start secreting its type of antibody. All its descendants—a clone—will produce that same, single type of antibody, and its concentration in the blood will rise.
However, the vast majority of B cells will never meet that special matching antigenic determinant and they simply die after circulating for a week or two.
Inject a complicated mixture of proteins into an animal or human and you will activate hundreds of different B cell clones in the spleen and lymph nodes, and each will produce its own type of specific antibody. All these antibodies will be present in blood and antiserum. This complexity can limit the usefulness of antibodies as analytical tools in science, and it is this problem that monoclonal antibodies circumvent.
Monoclonal antibodies are a brilliant invention that allow the production of just a single type of mouse antibody—the product of just a single clone, hence the name monoclonal in large amounts. How?
B cells are normal mortal cells, and after a certain number of cell divisions they die. However, cancer cells (or transformed cells as they are often called) are effectively immortal—they can be kept dividing in tissue culture as a cell line for ever. Imagine a transformed cell line that produced a useful antibody. You could get as much antibody as you wanted.
A monoclonal accomplishes this. Immortal B cell lines, derived from cancerous B cells—called myeloma cells—have been established from both mice and humans. A monoclonal antibody is the product of a normal antibody-producing mouse B cell fused with an immortal mouse myeloma cell.
A mouse is given a few immunising injections as usual with, say, membranes from human breast cancer cells. Once the immune response is well under way—three weeks—the mouse is killed and its spleen removed. Tens or hundreds of millions of actively growing antibody producing B cells will be concentrated in the spleen which is ground up to produce a suspension of cells. A special strain of cultured mouse myeloma cells is added to the suspension.
By adding warm polyethylene glycol to this mixture of cells, their membranes are destabilised and cell fusion takes place. Of course, the fusion is random: myeloma cells fuse with myeloma, spleen cell with spleen cell, but spleen-myeloma fusions also occur. These are what are wanted. In these fusions, immortality from the transformed cells will be transferred to an antibody-producing B cell from the spleen to produce an immortal, antibody-producing cell line. Unfortunately, a number of practical obstacles have to be overcome before this elegant idea is fully realised.
The whole mixture is transferred into hundreds of small wells in sterile plates so there are not too many thousand cells per well and cultured. What happens? The unfused spleen cells and spleen-spleen fusions naturally die within days—they are mortal. The spleenmyeloma fusions (which you want) grow—but so do the unfused myeloma cells and myeloma-myeloma fusions which are of no use.
To get rid of these, another trick is employed. The myeloma cell line used in the fusion is a mutant line deficient in a particular enzyme. Placing these cells in a special growth medium (HAT) that forces them to use that enzyme causes them to die. However, myeloma cell-spleen fusions survive in HAT because they are rescued by getting a functional gene from the spleen cell. Very ingenious! Hence, feeding the plates with HAT medium ensures that only spleen-myeloma cell fusions survive.
Beyond this there is a lot of work but it is not so clever. Culture fluids from all wells are screened for the presence of desirable antibodies. Once wells producing such antibodies are identified, cells from the wells are cloned by putting dilutions of suspended cells into hundreds more wells and then cells from antibody-producing wells will be cloned again. The aim is to get a clone of immortal, antibody secreting cells derived from a single cell. Months of intensive screening and recloning will be needed before a monoclonal cell line secreting antibody against an antigenic determinant on the protein of interest is finally obtained.
Herceptin is a monoclonal antibody made in this way. It binds to a protein that spans the membrane of human breast cells. This protein is a receptor to which a growth factor is supposed to bind. The protein is called HER2 (short for human epidermal growth factor receptor) or ErbB2. This growth factor—supposedly produced elsewhere in the body—should attach to the part of the protein receptor exposed on the outside of the cell, causing a change to the shape of the receptor, which signals the internal machinery of the cell to start the cell division process. Several similar receptors are known—HER1, HER2, HER3, and HER4—but the actual growth factor that binds to HER2 has not been identified and it is doubtful that it exists, whereas it has been identified for the other receptors.
In about 25 per cent of breast cancers, an abnormally large amount of HER2 protein is produced (or over-expressed), and this excess tends to cause the receptor protein to aggregate and enter a permanently switched-on configuration, in the absence of the growth factor. Being permanently switched on means the cell keeps getting messages to divide, making it cancerous. Some of the aggregates will be with other receptors from the HER family.
There are a number of types of breast cancer. That type associated with over-expression of HER2 is particularly aggressive and not very susceptible to treatment by standard chemotherapy drugs.
Herceptin is only used to treat those breast cancers where the HER2 receptor is overexpressed. How does it work to slow the cancer? Antibody-antigen complexes on a cell surface attract a cascade of blood proteins (termed the complement system) that together punch a hole through the cell membrane. This kills the cell. Antigen-antibody complexes on a cell surface also bind to receptors on immune system killing cells, which may then release compounds that kill the target cell. Binding of antibody to the receptor can also block the signalling system carrying the “divide” message to the interior, thus arresting the growth of the tumour. And blocking HER2 seems to inhibit the formation of new blood vessels nearby, a process essential for tumour growth. These are all good and useful actions for an anti-cancer drug.
However, Herceptin is still a mouse antibody. Inject it into humans and it will be seen as foreign and provoke the production of antibodies against itself. Yes, antibodies against antibodies. These human anti-mouse antibodies will bind to Herceptin, interfere with its anti-cancer activities and ensure that it is quickly taken out of circulation—not what you want for a particularly expensive drug.
This is why the humanising of the mouse antibody was undertaken. Human and mouse antibodies have a similar overall structure, shown on page 19. Each antibody consists of four chains of protein, two of them identical and longer (heavy chains) and the other two half the size, identical to each other, and termed light chains. The intact four-chain molecule has two identical antigen binding sites towards the ends of each pair of light and heavy chains.
Recall that antibodies produced by each clone will bind to different antigen shapes. This specificity is based on what amino acids are present in the antigen binding or CDR sites of different clones. While the antigen binding sites of antibodies are highly variable—and called V regions for short—the rest of the molecule is quite conserved and referred to as constant or C regions. Even within the V regions are more conserved framework stretches of amino acids which are liable to cry “mouse” to the human immune system.
Using the sophisticated techniques of genetic engineering, human heavy and light constant region genes were spliced into a new version of the gene in place of the mouse C regions. In the variable regions this could not be done because the binding specificity for HER2 had to be retained.
The sequences of amino acids in the V heavy (VH) and V light (VL) regions were determined (no trivial thing) and a detailed molecular model of the antigen binding region constructed. Human versions of the antigen binding region were then worked out with the aid of models. A series of short sections of human-sequence DNA (oligonucleotides) were then synthesised with a DNA synthesiser and joined to make a new heavy chain V region and a new light chain V region. The humanised VL required 24 amino acid changes from the original mouse, and the VH, 32 changes. This was all very complex and sophisticated genetic engineering. But there was still more.
Having engineered a human version of the mouse monoclonal that binds to HER2, could it be improved? With insights gained from the modelling programmes, a few amino acids were substituted in the CDRs (by altering the nucleotides coding for them) to see if the resulting antibodies bound the antigen more tightly than the original and if they inhibited the proliferation of human cancer cells overexpressing HER2 more effectively than the original. One of the seven variants produced, humAb4D5-8, bound to the target antigen threefold more tightly than the original mouse antibody, and 250-fold better than the straight humanised version which was—surprisingly—nowhere near as good as the mouse monoclonal. Furthermore, humAb4D5-8 supported antibody-dependent killing much more efficiently than either the mouse monoclonal or the original humanised version. And it failed to kill cells which expressed HER2 at normal levels—an excellent thing, because we assume that low levels of HER2 on normal breast cells are important.
It is not without interest to note that the original mouse monoclonal was made in the late 1980s, and that the genetic manipulation outlined here was described in a 1992 paper. In New Zealand 15 years later we are arguing about Herceptin as a new drug. Why has it taken so long to reach the shelves?
Two reasons I suspect. First, clinical trials to establish efficacy and safety take a very long time. Second, producing large amounts of the antibody may not be trivial. The scientists who engineered the antibody (staff from biotech company Genentech in San Francisco, now owned by Roche) originally worked with cultures that produced microgram (millionths of a gram) amounts of antibody per ml. The usual dose given to patients is 4 milligrams per kilo of body weight initially followed by 2 mg a week for up to a year. Genentech now produces the antibody in large-scale culture facilities, with the engineered gene that codes for the antibody inserted into Chinese hamster ovary cells. Production is likely elaborate, took a while to perfect, and quite expensive. (Early monoclonal antibodies for research use often cost US$1000 per milligram. Herceptin now costs $NZ3870 for a 440 mg vial, putting a regular maintenance treatment for a 65 kg woman at $1100).
Why is Pharmac dragging the chain?
Herceptin is still very expensive compared with most drugs, and this expense has raised questions for health services around the world, not just here in New Zealand. The reality for most health services is that if they spend tens of millions on Herceptin annually, they won’t be able to fund a whole lot of other things. So cost-benefit analyses are necessary here as in all facets of the health system.
How effective is Herceptin? Opinions are divided. Some medical researchers claim “the results are not evolutionary but revolutionary.” Others are more cautious. Unfortunately, Herceptin has been found to be associated with heart problems in up to 28 per cent of patients (more typically 3–7 per cent), especially when used in conjunction with a class of anti-cancer drugs know as anthracyclines which also cause heart damage. Some say the heart damage is modest and reversible once patients come off Herceptin (J Clin. Oncol. 2007, 25 (25) ). Others are not so sure that it is minor (J Clin. Oncol. 2007, 25 (23) p3525). At least one report claims that giving Herceptin before anthracycline chemotherapy for a short time is effective and causes no heart damage. HER2 proteins are thought to play an important role in the development of cardiac cells. The anthracycline drug Adriamycin, also known as doxorubicin, and an important ingredient in breast cancer chemotherapeutic cocktails, is known to damage cardiac cells, and Herceptin may block the development of replacement cells or repair in damaged cells.
Treatment duration is another issue. Initial trials of Herceptin used it for a year—or even two—although those were arbitrary periods. A much smaller, more recent Finnish trial used it for only nine weeks and claimed similar results. In New Zealand, Pharmac has approved Herceptin for nine-week treatment courses of early breast cancer (see below), much to the outrage of some women’s groups. Several studies have noted the rapid emergence of resistance to Herceptin in many patients. If true, prolonged treatment would be of little benefit.
What are treatment results like with Herceptin? In two large studies, the most recent monitoring four years after treatment finished revealed that 92.6 per cent of women treated with chemotherapy plus Herceptin were still alive and 85.9 per cent were cancer-free, compared with 89.4 per cent alive and 73.1 per cent cancer-free of those just treated with chemotherapy.
These results are described as a 52 per cent reduction in disease recurrence and a 35 per cent reduction in the risk of death associated with the use of Herceptin. Such accounting is somewhat misleading. The benefit from Herceptin treatment ranges from 0–9 per cent improvement in overall survival depending on the study. Looked at a different way, how many patients in trials needed to be treated with Herceptin to have one extra woman alive after two years? Two of the larger trials came up with 55.6 women, a third with 34.5, but the nine-week Finnish trial came in at a more impressive 14.5 women. Some deaths from Herceptin-related heart problems counterbalance the breast cancer patient lives saved. Comparisons between trials are complicated by factors such as, were patients who already had some heart disease excluded, which of numerous chemotherapy drugs were used in conjunction with Herceptin and were they used simultaneously, before or after and for how long. There are also variations in outcome depending on the stage of the cancer.
In advanced breast cancer where the disease has metastasized, the outcomes are different from in early breast cancer. In advanced disease, use of Herceptin in addition to chemotherapy results in:
—longer time to disease progression (7.4 vs 4.6 months)
—longer response to drugs (9.1 vs 6.1 months)
—fewer deaths at 12 months (22 vs 33 per cent)
—longer survival (25.1 vs 20.3 months).
The use of Herceptin in HER2 positive advanced breast cancer is well accepted. However, its use in the much larger pool of early breast cancer patients has provoked controversy, both in New Zealand and overseas. All concerns relate to the cost-benefit ratio.
For instance, in the Br. Med. J. 333, (2006), p1118–20, clinicians calculated that they would need £1.9 million to treat 75 patients with Herceptin. That would mean they could not afford to treat 355 patients receiving adjuvant chemotherapy treatment (usually chemo being given after surgery or radiotherapy)—where the cost-effectiveness is far better than that currently expected for Herceptin—and 16 of those patients would die. The use of Herceptin would likely save only one or two of the 75 treated.
British studies have found costs of £58,876 per quality adjusted life year gained (QALY). An Australian study has determined costs of $A22,793/ QALY, and $A414,012/ cancer death avoided for 52 week courses. One US study has ascertained $US34,201 per QALY gained, another $US39,982. A Norwegian study concluded that Herceptin was not cost effective for those with metastatic breast cancer. (Herceptin is priced differently in each market and is the main cost.)
Obviously, shorter courses of treatment would make the drug much more affordable, but it will likely be several years before results from full trials come in. There still look to be quite a lot of possible improvements to be gained from tinkering with treatment regimes involving Herceptin and different combinations and timings of chemotherapy drugs. What of the future? As DNA profiling of individual tumours becomes increasingly feasible, I suspect more
subtypes of breast cancer will be recognised and each will come to have its own treatment. The whole HER2-Herceptin story is an early step in this direction. A further step was taken in a paper earlier this year (http://jnci.oxfordjournals.org/cgi/content/full/99/9/694) in which Herceptin, another new monoclonal, Pertuzumab, against HER2 still in clinical trials, and a HER2 signalling inhibitor, Gefitinib, were all combined to treat experimental human breast cancers grown in mice. All three agents block growth signals from the HER1, HER2, HER3 and HER4 system in different ways. Pertuzumab binds to a different antigenic determinant on HER2 from Herceptin. When all agents were used together—but not any two—almost all tumours were eradicated as long as estrogen receptors were also blocked. Finally, Herceptin may find use in treating other cancers. HER2 may be involved in at least some ovarian and pancreatic cancers. We can also be sure that further clinically useful monoclonal antibodies will appear and they will all be expensive.