Pituitary tumours

suppressor gene silencing. The different clinically recognized types of pituitary tumour are currently treated by medical therapies such as dopamine and somatostatin agonists, surgery or radiotherapy. However, these treatments are not entirely satisfactory and recent advances in gene therapy may offer valuable new therapeutic opportunities for patients with aggressive tumours that fail to respond to traditional approaches.

Pituitary tumours are a common type of intracranial neoplasm and, depending on the cell type of origin, have diverse endocrine and reproductive effects. The developmental biology of the different cell types is understood to result from a sequential activation of a cascade of transcription factors, and mutations in these factors result in various forms of hypopituitarism. Tumours in the pituitary gland arise from activation of dominantly acting oncogenes such as gsp, or from loss of function of a series of tumour suppressor genes such as MEN1. Abnormal patterns of DNA methylation may be implicated in the allelic losses that cause tumour suppressor gene silencing. The different clinically recognized types of pituitary tumour are currently treated by medical therapies such as dopamine and somatostatin agonists, surgery or radiotherapy. However, these treatments are not entirely satisfactory and recent advances in gene therapy may offer valuable new therapeutic opportunities for patients with aggressive tumours that fail to respond to traditional approaches. et al., 1999). However, to date, no gain-of-function mutations have been reported in any of these transcription factors that might account for pituitary tumour development.

Incidence and prevalence of pituitary tumours
The peak incidence of pituitary tumours occurs in humans between the ages of 30 and 60 years, and occurs earlier in women (20-45 years) than in men (35-60 years) owing to the greater frequency of prolactinomas in young women (Faglia, 1993). Pituitary tumours are rare in people under 20 and are mostly prolactinomas (86%) or corticotrophinomas (10%) (Mindermann and Wilson, 1994).
Pituitary tumours occur as part of multiple endocrine neoplasia (MEN-1) in 30-40% of cases and are associated with loss of function of the tumour suppressor MEN-1 gene product. In MEN-1, prolactinomas and somatotrophinomas are the commonest tumour subtype. MEN-1 gene abnormalities are rare in sporadic pituitary tumours. Rarer still are familial cases of acromegaly that are not associated with MEN-1 and, in these cases, the genetic basis is unclear (Teh et al., 1998). Somatotrophinomas, specifically, may also occur as part of the familial Carney complex of cutaneous lentiginoses, atrial myxomas and other neoplasias (Asa et al., 1993).
When considering the prevalence of pituitary tumours, it is important to distinguish between clinically overt tumours and incidentally discovered cases. In a stable catchment population of about one million inhabitants approximate incidence and prevalence figures based on the 10 year period, 1988-1998, are shown (Table 1). These data include all patients investigated by endocrinologists, including those not referred for pituitary surgery. Incidence and prevalence figures derived from surgical series alone may be underestimates since some patients may not undergo pituitary surgery and may be treated primarily by different modalities, that is, radiotherapy, adrenalectomy or drugs.
The frequency of different pituitary tumour subtypes is shown (Table 2) from a surgical series (Mindermann and Wilson, 1994) of over 2000patients undergoing operation in USA between 1969and 1993. Prolactinoma is the commonest pituitary tumour subtype but, if surgical series from Europe are examined over this same time period, this subtype would be much less prevalent due to the primary treatment of prolactinomas with dopamine agonist drugs, which were not then available in the USA. This analysis illustrates the potential bias that can be introduced by the design of this type of study.
Pituitary tumours found incidentally as part of radiological investigation of the brain for unrelated symptoms are quite frequent owing to the widespread availability of computerized tomographic (CT) and magnetic resonance (MRI) imaging. However, before these imaging modalities were readily available, many autopsy series examined pituitary glands for adenomatous change. The frequency of adenomas in autopsy series varied from 1.5 to 27% with an average of 11.3% from a total of 12 411 pituitaries (Molitch, 1997). Tumours were equally divided between the sexes, and were distributed evenly across age groups (16-86 years). Only three of the 1403 adenomas found were > 10 mm in diameter. In autopsy material that has been examined immunohistochemically, 46% of adenomas stained for prolactin (Molitch, 1997).
A few studies have evaluated CT and MRI scans in randomly selected normal individuals. In these studies, only lesions > 3 mm in diameter were deemed to be reliably distinguishable from background image noise (Molitch, 1997). The frequency of CT-identified abnormal images (focal hypo-or hyper-density) varied from 3.7 to 20% in three studies (Molitch, 1997). With MRI, focal hypodensities of 2-5 mm in size were found in 20 of 52 normal subjects in non-enhanced images and 3 mm thickness slices (Chong et al., 1994), which appears to be a very high frequency (40%) of abnormalities. In another study of 100 volunteers (70 women and 30 men), focal hypodensities у 3 mm were found after gadolinium administration in seven women and three men, resulting in an overall radiological prevalence of pituitary abnormality of 10% in the asymptomatic population .

Pathogenesis of pituitary adenomas
The monoclonal nature of the majority of pituitary tumours was elucidated by X-chromosomal inactivation analysis in female patients (Alexander et al., 1990;Herman et al., 1990). In female somatic tissues, one of the two X chromosomes is inactivated (either the maternally-derived or the paternally-derived X chromosome). This inactivation event is assumed to occur randomly throughout normal tissues, resulting in a polyclonal mosaic. Thus, a tumour arising in adult life would be polyclonal if it arose from hyperplastic expansion from a population of neighbouring cells, but monoclonal if it arose from a single progenitor cell that had undergone somatic mutation. Clayton et al. (1999) confirmed this assumption by both loss of heterozygosity (LOH) analysis and X-chromosome inactivation, but also showed that recurrent tumours are frequently derived from separate, independent, clones from the initial tumour. This finding indicates that either more than one clone is present from the outset, although only one dominates, or that different clones are temporally separated. Although the study of Clayton et al. (1999) supports the view that established pituitary adenomas result from a clonal expansion of a single mutated pituitary cell, it is also possible that tumour initiation arises from a background of hyperplasia of the respective cell type. Several animal models have provided support for the concept of hormonal stimulation in the development of pituitary tumours that may occur on a background of preceding hyperplasia (for review, see Asa and Ezzat, 1998). The importance of hypothalamic factors, in some cases of local pituitary origin, and of peripheral hormones in influencing tumour growth and progression is supported by clinical findings (for reviews, see Herman-Bonert and Fagin, 1995;Asa and Ezzat, 1998).
The balance between growth and growth-inhibitory proteins that regulates progression through the cell cycle is normally tightly regulated. A cascade of products signal a cell to divide (proto-oncogenes), to remain quiescent (tumour suppressor genes (TSGs)), or to undergo terminal differentiation. Alterations in these pathways are the hallmark of tumorigenesis. Oncogenes are responsible for a gain of function and are dominant at the cell level and require only a single 'hit'. With the exception of the gsp oncogene, there have been only isolated reports describing the involvement of well-recognized oncogenes (fos, myc, ras, erb-A) in pituitary tumorigenesis. The principal mechanisms responsible for oncogene activation are mutation, amplification and translocation. In the case of the TSGs, tumorigenesis results from a loss of function that requires that both copies of the gene be inactivated, so two 'hits' are required.

Proteins with intrinsic GTPase activity
The G proteins comprise a family of heterotrimeric proteins that transduce signals from the cell surface via ligand-receptor interactions to downstream effectors. The α-subunit of these complexes (Gs, Gi and Gq), in their active GTP-bound form, represents the proximal effector molecule in these signalling cascades. Heterozygous  activating somatic point mutations in the α-subunit of the stimulatory Gs protein (Gα s ) were the earliest dominant activating mutations described in pituitary tumours (Landis et al., 1989;Farrell and Clayton, 1998 and references therein). Activation results when mis-sense mutations replace residues 201 (Arg to either Cys or His) or 227 (Gln to Arg or Leu). The resultant oncogene, termed gsp, occurs predominantly in GH-secreting tumours. Gsp mutations are rare (< 10%) in other pituitary tumour subtypes. No consitent association between the gsp oncogene and clinical features has been shown. The direct mechanism by which the ensuing increased cAMP concentrations stimulate somatotroph proliferation and GH hypersecretion is not known but may involve the phosphorylated (activated) form of the transcription factor cyclic AMP response element binding protein (CREB). The analogous activator of the phospholipase Cβ-Ca 2+protein kinase C pathway, Gα q , has been screened for activating mutations in nucleotides corresponding to those of Gα s , and none have been found.

Pituitary tumour transforming gene
Over-expression of a novel pituitary tumour transforming gene (PTTG) in NIH3T3 cells was shown to result in cell transformation in vitro and to induce tumours in vivo in athymic mice (Pei and Melmed, 1997). Moreover, increased amounts of PTTG mRNA were detected in all subtypes of human pituitary tumours, with the largest tumours exhibiting a greater than tenfold increase (Zhang et al., 1999a). The PTTG protein contains a SH3-docking domain, indicating its involvement in intracellular signalling (Zhang et al., 1999b) and potently induces the expression of basic fibroblast growth factor (bFGF), a mediator of cell growth and angiogenesis. The expression of PTTG in most pituitary tumours studied indicates that this is an early change in pituitary tumorigenesis and may act in a paracrine fashion to induce bFGF.
Mechanisms responsible for loss of tumour suppressor gene function There are three principal mechanisms for loss of function of a TSG: (i) Loss of function may result from loss of one allele (LOH) accompanied by mutation in the retained allele. The prototype example of this is the retinoblastoma gene (RB1) in both familial and sporadic forms of retinoblasta (Knudson, 1971). (ii) Homozygous deletion of both alleles of a gene may also lead to loss of function of a TSG. A good example is p16/CDKN2A in head and neck cancers, in which 70% of primary tumours lose the gene through this mechanism (Reed et al., 1996). (iii) Loss or reduced expression of a TSG may occur through methylation of associated cytosine-guanine dinucleotide (CpG) islands. For the TSG p16/CDKN2A gene, loss of protein through methylation approaches 40% in primary colon tumours and increases to 92% in their equivalent cell lines (Herman-Bonert and Fagin, 1995).
However, analysis of chromosome 9p has marked this region as an early target in pituitary tumorigenesis with an LOH frequency of 31% in both invasive and non-invasive non-functional tumours, indicating this to be an early change in pituitary tumorigenesis (Farrell et al., 1997). Most tumours showed non-contiguous deletions that flanked, but excluded loss of, the p16/CDKN2A gene itself.
Characterized tumour suppressors associated with regions of allelic loss MEN1. Multiple endocrine neoplasia type 1 (MEN-1) is an autosomal dominant endocrine cancer syndrome consisting of parathyroid, pancreatic islet, and pituitary tumours. Germ-line mutations, leading to loss of function, have been identified in both familial and sporadic MEN1 patients (Agarwal et al., 1997;Farrell et al., 1999 and references therein). Mutations in the retained MEN-1 allele are frequent in tumours from MEN-1 patients. Inactivating mutations in the MEN-1 gene are extremely rare in sporadic pituitary tumours and examination of menin gene transcript expression failed to identify any tumours showing reduced mRNA expression (Farrell et al., 1999), although alterations in post-translational mechanisms leading to a reduction in menin protein expression have not been excluded.
Retinoblastoma-susceptibility gene. The findings in RB1 transgenic knockout mice of pituitary adenocarcinomas of intermediate lobe origin prompted study of LOH at the RB1 locus. Pei et al. (1995) failed to detect loss of retinoblastoma protein (pRB) by immunohistochemistry in particularly aggressive adenomas or metastatic pituitary carcinomas, despite these tumours sustaining LOH at an intragenic marker. Bates et al. (1997) found that a small proportion of tumours failed to express pRB, although loss of protein expression was not associated with LOH of an RB1 intragenic marker. Simpson et al. (1999a) showed that 27% of somatotrophinomas, compared with 4% of non-functional tumours, failed to express pRB, but this did not correlate with LOH of an RB1 intragenic marker. Thus, chromosome 13q LOH may signify another TSG in the vicinity of RB1.
p53. The p53 gene is the most commonly altered locus in human cancers. Single strand conformational polymorphism (SSCP) and direct sequence analysis (Levy et al., 1994) failed to detect mutations in this gene, leading to the suggestion that it is not implicated in pituitary tumori-genesis. However, other reports have highlighted the lack of concordance between detectable p53 protein and a corresponding gene mutation. Although some studies failed to detect immunohistochemical expression of p53 protein in pituitary adenomas, others have detected immunopositivity associated with invasive non-functional tumours and corticotrophinomas (Buckley et al., 1994;Thapar et al., 1996). The significance of these findings with respect to p53 gene regulation in pituitary tumours is unclear.
p27. Transgenic mice lacking the TSG p27 develop multi-organ tumours that include those of pituitary intermediate lobe origin (Nakayama et al., 1996). Tanaka et al. (1997) failed to detect loss of microsatellite markers in the region of this gene on chromosome 12p12-13 or changes at the nucleotide level. Immunohistochemical analysis of p27 protein showed that the highest concentrations were found in non-tumorous pituitaries, with a significant decrease in the number of cells expressing p27 during progression from normal pituitary to pituitary adenoma and carcinomas (Dahia et al., 1998). Downregulation of p27 protein is thought to occur through a post-translational mechanism, implicating the ubiquitin-proteosome pathway in this process.
p16/CDKN2A. The CDK inhibitor p16 is recognized as a tumour suppressor associated with numerous tumour types. As a cell cycle regulator, by binding or sequestration of CDK4, it prevents the phosphorylation of pRB and is responsible for inhibiting progress through the G 1 /S cell cycle checkpoint. Loss of p16/CDKN2A, by several mechanisms, results in pRB remaining in its hyperphosphorylated form, negating its ability to inhibit progress through this cell cycle checkpoint. As already indicated, the p16 gene is neither homo-or hemizygously deleted in pituitary tumours. SSCP analysis and sequencing of variants in a series of 31 sporadic pituitary tumours also failed to detect mutation in either the p16/CDKN2A or the related p15/CDKN2B gene on chromosome 9p21 (Yoshimoto et al., 1997).
Therefore, the methylation status of the CpG island associated with the p16/CDKN2A gene has been studied, complemented with assessment of protein expression by either western blot analysis or immunohistochemical expression (IHC). p16 protein is not detected or is barely detectable in most tumours that are methylated (Woloschak et al., 1997;Simpson et al., 1999b). Moreover, subdivision of the non-functional tumours into invasive and noninvasive cohorts showed an approximately equal frequency of methylation in each, indicating that this epigenic phenomenon represents an early change in pituitary tumorigenesis (Simpson et al., 1999b).

Methylation: a unifying mechanism?
The combined data from numerous reports of human pituitary tumours that used an LOH approach demonstrate losses associated with early changes in pituitary tumorigenesis and in the transition from the non-invasive to the invasive and metastatic phenotype. Although LOH is assumed to define DNA regions harbouring putative TSG loci, studies of the retained allele associated with these regions of loss have revealed infrequent mutation in known TSGs. Neither have consistent mechanisms been described for the reduced expression of other TSGs associated with pituitary tumour progression. However, there is evidence from tumours of neural, renal and colonic origin that abnormal methylation may precede and predispose toward genetic instability and lead to loss of genetic material. Moreover, the progressive increase in CpG island methylation may be associated with progressive TSG silencing (Jones, 1996). If this is the case, allelic losses described in pituitary tumours may be a consequence of a preceding abnormal methylation pattern. Methylation would initiate the allelic loss, although this would not necessarily encompass either putative or known TSGs.

Transgenic model systems
In addition to the transgenic models already discussed (RB1 and p27), data on targeted expression of pituitary-specific growth factors have been reported and may serve as model systems for growth factor-induced tumours. McAndrews et al. (1995) described a transgenic mouse harbouring a lactotroph-targeted transgene, in which the transgene over-expressed transforming growth factor α (TGFα). The transgenic mouse demonstrated selective pituitary lactotroph hyperplasia and formation of a prolactin-containing adenoma. In another transgenic model, over-expression of pituitary-directed nerve growth factor (NGF) gene resulted in lactotroph hyperplasia (Borrelli et al., 1992). Studies using transgenic mice that over-expressed the GHRH gene showed that prolonged stimulation eventually resulted in pituitary tumours but that this was preceded by somatotroph hyperplasia (Asa et al., 1992). Although transgenic mouse models represent important findings, to date they have not translated into meaningful data with regard to the ontogeny or behaviour of human pituitary tumours.

Clinical aspects
Pituitary tumours are classified according to their patterns of hormone production and their clinical presentation. Tumours are described as 'functioning' if they over-produce prolactin, growth hormone (GH), adrenocortocotrophic hormone (ACTH), gonadotrophins or thyroid-stimulating hormone (TSH), or 'non-functioning', if no clinical syndrome is apparent, although some of these tumours produce gonadotrophin subunits.
All subtypes of tumours may present as microadenomas, defined as tumours contained within the pituitary fossa, and < 1 cm in diameter, or as macroadenomas that have extended outside the fossa to invade adjacent tissues (from 1 to > 10 cm in diameter). Tumour extension outside the pituitary fossa may involve the cavernous sinus laterally (parasellar extension), the optic chiasm superiorly (suprasellar extension), or the sphenoid sinus inferiorly. Rarely, aggressive tumours may extend along the skull base, even sometimes causing multiple cranial nerve palsies (for example, see Davis et al., 1990), but only exceptionally are they truly malignant with distant metastasis. Macroadenomas commonly come to clinical attention through pressure effects such as headache or progressive visual failure, whereas microadenomas are usually identified through investigation of a clinical endocrine syndrome, or may be discovered as incidental findings during scanning of the brain for other reasons (see above).
Lactotroph tumours (prolactinomas) are the commonest type of functioning pituitary tumour, and the prevalence of hyperprolactinaemia in the population may be as high as 0.5% (Miyai et al., 1986). The hyperprolactinaemia caused by a pituitary adenoma needs to be distinguished from hyperprolactinaemia caused by, among other things, pregnancy, drugs (such as dopamine antagonists and phenothiazines), hypothyroidism and polycystic ovary syndrome. The hyperprolactinaemia has direct effects, for example, on the breast to cause lactation, and indirect effects on gonadotrophin secretion that lead to hypogonadism. In general, prolactinomas come to medical attention in women because of the syndrome of galactorrhoea (lactation) and amenorrhoea, whereas in men lactation is very rare and the symptoms of impotence and lack of libido tend to come to medical attention relatively late, which may be one reason why men tend to present with larger tumours than women, although there may also be biological differences in tumour behaviour.
Non-functioning tumours comprise 40% of all pituitary tumours, and perhaps because they produce no clinical syndrome apart from eventual hypopituitarism, they tend to be large by the time of presentation, and frequently cause headache and visual impairment before they are diagnosed.
Somatotroph tumours secreting GH cause the clinical syndrome of acromegaly, affecting about 40 individuals per million in Western Europe. The GH excess stimulates high concentrations of insulin-like growth factor I (IGF-I) production by the liver, and the condition results in overgrowth of many tissues. The diagnosis is often delayed by many years, as the changes in soft tissues occur slowly, and perhaps for this reason tumours are commonly larger than 5 mm in diameter, and larger tumours (> 10-15 mm) may also give mass effects such as visual impairment by the time they are diagnosed (Duncan and Wass, 1999).
Corticotroph tumours secreting ACTH and other proopiomelanocortin (POMC)-derived precursor peptides, are rare and cause the clinical syndrome of Cushing's disease with ACTH-dependent adrenal hyperplasia and hypersecretion of glucocorticoids. The syndrome of glucocorticoid excess is often marked and may be life-threatening, resulting in obesity, high blood pressure, osteoporosis, depression, thinning of the skin and poor wound healing. Nonetheless, the pituitary tumours are frequently small and may not be visible even on high-resolution magnetic resonance scanning.

Current therapy for pituitary tumours
The aims of any therapy for pituitary tumours depend on the clinical presentation but, in general, the aim of treatment will be to reduce tumour mass and hormone excess. Debulking of significant tumour mass is important to reduce symptoms of headache resulting from compression of surrounding structures, and particularly to relieve pressure on the optic nerve from suprasellar extension. However, normalization of the endocrine abnormality is coming to be seen as equally important for long-term health and even longevity, and that, therefore, an ideal therapy should also reduce hormone hypersecretion to normal or 'safe' amounts (for discussion relating to acromegaly, see Bates et al., 1993;Sheppard, 1994;Orme et al., 1998).
Surgery offers the potential for long-term cure by total excision of a pituitary adenoma, leaving intact the remaining normal pituitary gland. In the treatment of microadenomas, trans-sphenoidal pituitary microadenomectomy is safe and carries low risks of damage to the normal pituitary. Indeed, recovery of pituitary function may occur after successful resection of pituitary macroadenomas by transphenoidal surgery (Arafah et al., 1994). Nonetheless, in a series of patients with GH-secreting tumours operated by a specialist surgeon, 14% developed new pituitary dysfunction as a result of the surgery (Ahmed et al., 1999). Surgery for macroadenomas is often essential for debulking and relief of pressure symptoms, but as the tumours are larger, there is often a greater risk of damage to related structures. Nearly all tumours can be treated by the trans-sphenoidal route, but some require transfrontal craniotomy, which carries greater operative risks.
However, despite a generally satisfactory safety record, surgery carries surprisingly low cure rates for the endocrine abnormality, and the results are poorer the larger the tumour. Published surgical series generally reflect the results of highly specialized centres, but even taking this into account, the cure rates remain disappointingly low (Soule et al., 1996;Molitch et al., 1997). In summary, surgery has potential for cure and therefore remains a first choice of therapy in many circumstances, but the actual outcomes are often disappointing, even in the series of specialists. No pituitary tumour, however large, should be operated on without prior ascertainment of serum prolactin, since even large macroprolactinomas frequently shrink rapidly with dopamine agonist therapy (Davis et al., 1990;Bevan et al., 1992).
Medical therapy has brought about a revolution in expectations of treatment for pituitary disease. In many cases, drug treatment alone is adequate without recourse to any other therapy. Key medical therapies include dopamine agonists for hyperprolactinaemia, and somatostatin analogues and GH antagonists for acromegaly.
The astonishing success of dopamine agonists in both reducing serum prolactin and causing shrinkage of prolactinomas has allowed drugs such as bromocriptine, cabergoline and quinagolide to be used as the sole therapy for many patients with hyperprolactinaemia (Bevan et al., 1992). However, despite the success of dopamine agonists in 85-90% of patients, there are often significant predictable dopaminergic side-effects, notably nausea and vomiting, postural hypotension and dizziness, headache and constipation. Depressive reactions may also arise in some patients.
Owing to these side-effects, though remarkably effective, dopamine agonists are often disliked by patients and sometimes cannot be used. If treatment with dopamine agonists is required for fertility, it is usual to stop this as soon as pregnancy is confirmed. Should a prolactinoma enlarge during pregnancy, treatment with bromocriptine is safe. Somatostatin analogues do not cause such marked shrinkage of somatotroph tumours, but they frequently reduce GH, if not to normal, at least to concentrations that appear to be 'safe' in terms of normalizing long-term mortality (Sheppard, 1994). The clinical use of somatostatin analogues has been much eased by the introduction of depot preparations of octreotide and lanreotide that can be given at 2-4 week intervals, but there is still a requirement for uncomfortable repeated injections over long periods of time, and treatment can involve great expense over many years. GH antagonists are under clinical trial at present, and may represent an important advance, although they are unlikely to affect the size or growth of the underlying pituitary tumour (Thorner et al., 1999). GH antagonists appear to be well tolerated, although information on sideeffects is likely to emerge from current trials. For small tumours, both somatostatin analogues and GH antagonists may find a place as the sole therapy for acromegaly. However, the size of the pituitary tumour frequently necessitates surgery since, in contrast to dopamine agonists for macroprolactinomas, somatostatin analogues produce, at best, a small decrease in pituitary size, and this occurs only slowly over several months.
Pituitary irradiation is effective at reducing tumour growth in the long term, and is often used as an adjunct to surgery when this has failed to achieve an adequate cure. Although irradiation is proven to reduce the risk of postoperative tumour progression, this is almost invariably at the cost of eventual hypopituitarism (Littley et al., 1991), which then requires life-long multiple-hormone replacement therapy. Thus, patients require regular screening for progressive hypopituitarism, and eventually need substitution therapy with corticosteroids, GH, thyroxine and sex steroids (or gonadotrophins to achieve fertility).
In summary, at present, despite important advances, pituitary tumour therapy is often unsatisfactory. Surgery is commonly inadequate to achieve endocrine cure, medical therapies have significant long-term side-effects and expense, and irradiation causes hypopituitarism, which itself may result in reduced life expectancy so is to be avoided if possible. Therefore, it is timely to consider whether recent advances in pituitary cell and molecular biology can be used to design new therapies, with the aim of selective ablation of tumour cells, without damaging the normal pituitary gland.

Gene therapy?
Now there is a substantial background of knowledge concerning the regulation of pituitary hormone gene expression, there is a strong case for applying this information to develop new tools for therapy. Recombinant adenoviruses deliver viral transgenes to pituitary cells efficiently in vitro (Castro et al., 1997) and, in principle, the tight transcriptional regulation of the prolactin gene could be exploited to direct expression of a desired transgene to lactotrophic cells only within the mixed cell population found in the intact pituitary gland. Appropriate expression of a marker gene can be effectively limited to lactotrophic cells (Southgate et al., 2000;Davis et al., 2001) such that, in principle, a 'suicide' gene could also be expressed in a celltype-specific manner to ablate lactotrophic cells while leaving other cell types unaffected. A similar argument can be applied to the use of the GH promoter to target somatotrophic cells (Lee et al., 1999). The use of regulatable promoter elements such as tetracycline-inducible vectors also allows transgenes to be activated only when desired, further limiting the expression of transgenes (Harding et al., 1998). In the future, a wide choice of possible strategies might be opened up: for example, effective targeting of transgenes to pituitary cells could be used to replace TSGs in tumours that have arisen as a result of TSG mutation, to deliver dominant negative antagonists to oncogenes, or to alter DNA methylation status in selected cell types. Great progress has been made with gene therapy strategies in cancer treatment and, given the background of better understanding of pituitary tumorigenesis, some of this knowledge may be applicable to new therapeutic developments for pituitary disease (Castro, 1999). Thus, there is a powerful case for developing a proof of the concept using pituitary hormone gene promoter-directed gene therapy in the context of wellvalidated cell culture systems and whole animal physiology, although it may be some time before new therapies are available to patients (for review, see Davis et al., 1999).