A sinister side to a synthetic sex hormone

Tales of the 'sexual revolution' of the 1960s might suggest that understanding of sex hormones is a relatively recent chapter in history. But this is wrong by at least 4000 years.

The ancient Sumerians practised castration in the 21^st century BC, as did many subsequent civilisations. These societies valued eunuchs as royal servants and administrators because they were incapable of having children and so were thought unlikely to be tempted to seize power from their masters.

An understanding that female sexuality was similarly controlled by the presence or absence of key organs came much later but had profound implications for women's ability to control their fertility. However, initial discoveries only hinted at sex hormones' potential, and there was a disaster looming that still hangs over some thousands of women today. This is the story of that disaster.

In 1900 gynaecologist Emil Knauer confirmed that removing the ovaries from mammals after puberty inhibits oestrus, the periodic cycle of intense mating urge. Knauer also found that as re-implanting these removed organs elsewhere in the body restored oestrus, it seemed that chemical messengers from the ovaries might be crucial to mammalian sexuality.

Knauer's results suggested that hormone replacement therapy might remedy menopausal disorders and a mixture of hormones extracted with ether from abattoir-sourced ovarian tissue was marketed from 1913 with variable benefits.¹

For chemists the synthesis of female sex hormones was a worthwhile challenge. If they could isolate and identify the structure of the pure hormone(s), then the hormones (and their congeners) might be synthesised to produce more potent, and cheaper, drugs. The problem was that the natural hormones were present in such tiny amounts that isolating enough for classical chemical investigations was impossible. 

Two breakthroughs in the 1920s and 1930s made progress possible. Edward Doisy examined vaginal smears from pre-pubertal mice and castrated female mice (or rats) under the microscope. When the animals were treated with a sex hormone, the cells of the smears showed changes associated with oestrus, providing a quick and relatively simple bioassay for these hormones.

The second breakthrough was by Selman Aschheim and Bernard Zondek, who used a similar pre-pubertal mouse model to devise a urine-based pregnancy test.¹ They suggested that pregnancy might be associated with a change in the hormone profile of the urine. Indeed it was, and urine became a convenient source of female hormones, although this was soon superseded by mares' urine, which was especially rich in these hormones - during pregnancy a mare would produce some 30g of sex hormones, about 10 times the human equivalent.

Now, with a ready supply of source material, structural studies could start. Most studies focused on oestrone, a ketonic metabolite of primary oestrogens that is also hormonally active. Its molecular formula, C₁₈H₂₂O₂, was consistent with a four-ring structure and pointed to a possible connection with the cholesterol-like molecules whose structure was being investigated intensively at the time.

The correct structure for cholesterol was announced in 1932 and in the same year two groups of workers proposed a structure for oestrone, although it took another three years to confirm it as correct.²

While synthesis of natural oestrone was some 16 years away, it seemed a realisable goal as early as 1933 and J. W. Cook, E. C. Dodds and C. L. Hewett tried using keto-tetrahydrophenanthrene (1) as an intermediate. 

They were unsuccessful, but as part of their studies they examined the oestrus-exciting activity of this relatively simple molecule using Edward Doisy's methodology. They reported, presciently:

The observations show that (it) is capable when injected into castrated animals of inducing oestrus exactly similar to that obtained by the injection of oestrogens. There are grounds for hoping that substances of a much higher order of activity will be found before very long.³

The Doisy test was cheap and permitted rapid screening. Within a month Cook and Dodds reported that the potent carcinogen, 1,2-benzpyrene, showed oestrus-inducing activity. Dodds followed up this work, and in 1936 published a list of eight para-hydroxy-substituted aromatic species that showed oestrus-inducing activity. Di-(p-hydroxyphenyl) dimethyl methane (2) was typical and a dose of 100mg precipitated oestrus in a rat.⁴ However, 100mg corresponded to a "human equivalent dose" of about 17.5g, too large for therapeutic use.

In April the following year, Dodds reported that not only did stilbene derivatives (e.g. 3) show a 10-fold greater activity (allowing doses in rats of 10mg), but a species based on half a stilbene-like molecule showed an astonishing oestrus-inducing effect.

Just one microgram of p-propenylphenol (also known as anol, 4) seemed as effective as oestrone itself. Dodds enthused:

It is as yet too early to discuss the...therapeutic application of this observation, but the fact that the potency is so high brings a new importance to the investigation of synthetic oestragenic agents.⁵

And so it did, with researchers across the world attempting to replicate the work before launching off into their own investigations. Writing again in Nature, just nine weeks later, Dodds continues the story:

Concerning the high oestrogenic activity of anol, we have received information from some workers who have confirmed our observations, and from some others who have been unable to demonstrate activity with doses very much greater than those described by ourselves. 

Berlin chemist Walter Schoeller investigated this apparent paradox and found that when "active" anol was recrystallised from chloroform, the activity remained in the mother liquor, rather than in the crystals of anol that were precipitated. At this early stage Dodds was unable to identify the component in the liquor, but hazarded a guess that because anol polymerises with ease there was a possibility he had been working with a polymer of anol, and that this had been responsible for the results he reported two months earlier. 

By the following January, Dodds had identified the mystery anol-derived compound as 'di-anol' (5), a dimeride of p-propenylphenol (4).⁶

The focus of the investigation then moved to Robert Robinson's laboratories at Oxford. Robinson found even more potent relatives of Dodds' di-anol and established their structural relationship to oestrone. 4,4'-Dihydroxy-α, β-diethylstilbene (6), was Robinson's major discovery, which he named stilboestrol (also known as diethylstilboestrol, DES) and drew (Fig 1) to emphasise its connection to oestrone.⁷

To synthesise DES (Scheme 1), Robinson reacted anisaldehyde - a very cheap precursor - in a process that could be scaled up to give kilogram quantities. In contrast to most current drug research, but in accord with the policy of the Medical Research Council that funded his project, patents were not taken out on the compound or the process used to make it. 

Because of the ready availability of cheap precursors and the lack of patent protection, DES was relatively cheap to make. As DES was active in microgram amounts, it comes as no surprise that from 1938 its therapeutic potential was investigated intensively. This research largely, but not exclusively, focused on 'female' complaints. A chronology of some of the applications is given in Table 1.⁸

Of all the treatments listed in Table 1, it was the use of DES for preventing miscarriage that was to have disastrous consequences. It was once thought that miscarriages might be caused by a deficiency of oestrogens. Using DES to reverse this seemed logical, and its use became widespread even before finally being approved by the US Food and Drug Administration (FDA) in 1947.  

Six years later the first trial of its effectiveness found that DES had no benefit over a placebo for preventing miscarriages, but it remained heavily promoted and widely used. Anxious parents, faced with a threatened miscarriage, perhaps felt that any approved (and thus presumptively safe) treatment was worth trying, even if its benefits were unclear. 

A cloud on the horizon

However, all this changed with a 1971 report in the New England Journal of Medicine.⁹ Although the rare cancer cervicovaginal clear-cell adenocarcinoma was almost always confined to elderly women, this paper reported it occurring in eight girls and young women between the ages of 14-22. Crucially, at least seven of these women's mothers had taken DES during pregnancy. Could this treatment have had: An adverse effect on the mothers themselves, but of sufficiently rare occurrence that it was not readily spotted; or a delayed sinister effect on the daughters of DES-treated mothers; and it if was, were the sons also at risk?

Such reports alone do not prove a link, but they raise crucial questions. The next step was to match a large group of women who had taken DES during pregnancy with a similar sized group who had not, and to compare the outcomes. In 1978 just such a study was published, comparing 693 women who had taken DES with 668 controls. Thirty-two of the treated women had developed breast cancer, compared to 21 of the controls, almost a 50 per cent increase (somewhat higher than found in later larger studies). A 1995 literature review came to the conservative conclusion that: 

Most evidence suggests that the increased risk for breast cancer among DES mothers is real but small, with less than a two-fold increased risk. The risk for no other cancer has been shown to be significantly elevated in DES mothers, but these risks have not been adequately explored.¹⁰

And the daughters?

Abnormalities of the reproductive system and fertility problems are relatively common among the daughters of women who took DES during pregnancy. These daughters also have a significantly higher risk of breast cancer, with two studies showing 90 per cent and 250 per cent increases in risk for individuals over 40 years of age.

The present recommendation is that daughters of women who took diethylstilboestrol during pregnancy should have lifelong screening and should probably avoid hormone-based contraceptive methods. Daughters also have a 40-fold increased risk of cervicovaginal clear-cell adenocarcinoma, but this is still a rare condition, so the individual risk is about 1 in 1000 for women under 35 years of age. There is concern that this risk may increase greatly when 'DES daughters' reach the age of about 70, the peak age of onset of adenocarcinoma in women who have not been exposed to DES.

And the sons?

For 'DES sons' the health statistics are less definitive and earlier concerns about higher rates of testicular cancer in DES-exposed sons are currently disputed. At present there is no connection between maternal DES exposure and prostate cancer in the sons but women who had DES therapy in the 1940s and 50s now have sons in their seventies and sixties - the age range associated with prostatic cancer. It will be important to see if the DES sons show an increased susceptibility to this disease.¹¹ However, a 1995 study found that the sons of DES daughters had a higher rate of malformation of their penis - in which the urethra ends on the lower surface of the penis rather than at the tip - with rates about five times that of the general population.¹²

The mechanism by which DES could cause effects on a third generation, who had never been directly exposed to DES, is puzzling. It is probably not genetic - the sons of men exposed to DES in utero do not have a greater risk of this genital malformation than the general population, unlike the sons of mothers who were exposed. Perhaps the changes in the reproductive systems of DES daughters lead to the changes, or perhaps there are enduring changes in hormonal systems that affect some developing boys. Some animal studies suggest there may be some epigenetic changes involved, but the situation remains unclear.

However, the concept that a drug given to one person can have a major effect on their future grandchildren is a terrifying prospect for drug developers, doctors and regulatory agencies. Many modern drugs come and go within 20 years or less, and DES was essentially gone within 40 years. The idea that important side effects may not be evident until a generation later raises major challenges. Animal studies offer one way of screening for generational effects, but when the particular issue is both rare and (at the time) unknown, it can very easily be missed, or just not occur in that relatively small number of animals. 

DES had the potential to be a horrendous human tragedy that would have eclipsed even the thalidomide disaster. It remains a disaster for those who have developed cancer, and a significant challenge for those facing lifelong monitoring.

Some five million American women - and millions around the world - took DES and approximately 5000 daughters in the US can expect to develop related adenocarcinomas.¹ But it could have been far worse, had the initial observations on the eight young women not been followed up.

Medicine safety relies on acute observation and systematic investigation of 'signals' of potential concern. At present, it offers the main defence against another DES disaster, though improving understanding of drug effects and genetic influences will complement this in the future.

Pete Ellis is professor of psychological medicine at the School of Medicine and Health Sciences, University of Otago, Wellington, New Zealand. Alan Dronsfield is emeritus professor of the history of science at the University of Derby. Alan-Shaun Wilkinson is a senior lecturer in the department of education, health and sciences at the University of Derby.