38 Paraneoplastic disorders and neuroimmunology

The extraordinary expansion in the field of neuroimmunology witnessed in the last decade is not just in the number of neurological disorders now considered to have an immune basis, nor the depth of understanding of disorders long known to be ‘neuroimmune’. Nor is it in the number of antibodies discovered and now testable, nor in the range of new immune suppressant or modifying treatments now emerging or already available. It is of course all of these things, but it is also more than the sum of these parts. What we are currently privileged to witness is the coming together of immunological understanding, the neurobiology of disease, and rational immune therapy, or at least the beginning of this process. To take one isolated example, neurogenetics and neurophysiology taught us about the clinical consequences of channel disruption; laboratory-based neuroimmunology showed antibodies to be capable of producing comparable acquired disease; and it seems likely that specific anti-B-cell humanized monoclonal antibodies offer the therapeutic potential to remove these channel-disrupting antibodies. Neither of these steps could be described in the last edition, and one can imagine similar dramatic changes will emerge before the next.

Immunology is at least in part a study of mechanisms of directed molecular and cellular injury, which has as one major and ultimate result damage to a specific target. To this end a variety of cytotoxic effectors may be recruited; the complex network of cells and soluble molecules that otherwise constitute the immune system is responsible for initiating, activating, coordinating, targeting, restraining, recording, and reproducing this response.

The conventional but perhaps oversimplified division into cellular and humoral immune responses remains a helpful way of classifying mechanisms of immune activity, and providing a structure within which to explore components and effectors, and of course this particularly applies to neurology, where we think of a number of the central nervous system (Section 38.3) and particularly peripheral nervous system disorders, with myasthenia as perhaps the prototypic disease, as being B-cell-mediated. In truth, however, while this division remains convenient, it is not sustainable to pigeon-hole either immune reactions or autoimmune diseases as ‘T-cell-’ or ‘B-cell-’ mediated—the two arms of the immune response are wholly and inextricably interdependent.

This depends on circulating antibodies which are immunoglobulins secreted by plasma cells following antigenic stimulation of B-lymphocytes. The combination of antibody with antigen may act to neutralize the target, recruit and activate macrophages, or bind and activate complement.

These are cells of the macrophage/monocyte lineage, and are present in most tissues, including the brain, where they are represented as microglia. They have a key role both in the initiation of immune responses (Section 38.2.2) and as effector cells. Macrophages are able to attack their target either by phagocytosis or by secreting a variety of toxic substances, including free oxygen radicals and tumour necrosis factor. They also have a role in immune regulation and secreting proteins and peptides, macrokines, which act locally and influence other immune cells (Block 2005).

It is a group of serum proteins which circulate in an inactive form; activation is initiated classically by antibody, but also by an alternative pathway, and results in the formation of vasoactive and chemotactic peptides and terminal membrane attack complexes, which may injure or lyse target cells.

are centred primarily around T-lymphocytes and these also respond very specifically to individual antigens; in fact, the surface T-cell antigen receptor which binds antigen is structurally related to immunoglobulin, and both derive from the same gene superfamily. All T-cell antigen receptors are dimeric complexes of either an α- and a β-chain, or a g- and a d-chain. Antigen specificity of T-cell antigen receptors is a function of these chains’ tertiary structure; diversity is generated mostly by random recombination events affecting those groups of genes coding for peptides within the chains—closely reflecting mechanisms of immunoglobulin generation. Different T-cell antigen receptor can recognize the same epitope, and different epitopes may interact with the same T-cell antigen receptor.

The binding of antigen alone to T-lymphocytes is not, however, sufficient to activate the T-cell; an additional restriction is provided by the necessity for the antigen to be presented to the T-cell after antigen ‘processing’, which requires enzymatic breakdown of antigen to an appropriate form of 9–15 amino acids in length, and which can be effected only by Antigen Presenting Cells and also as a complex with a HLA molecule encoded by the major histocompatibility complex.

Class II major histocompatibility complex molecules, HLA-DR, are constitutively expressed only by antigen presenting macrophages and dendritic cells. Other cell types may be induced by lymphokines to express Class II antigens and so can acquire a role in T-cell stimulation; such cells are known as non-professional antigen presenting cells. Furthermore, even the formation of the trimolecular complex consisting of antigen, HLA-DR molecule, and T-cell receptor, is insufficient to allow T-cells to generate an antigen-specific response; binding is also required between a number of accessory molecules expressed by the antigen presenting cells and T-cell, including CD3 and either CD4 or CD8. Additionally, co-stimulatory signals, B7-1 and B7-2, provided exclusively by ‘professional’ antigen presenting cells, must also be present.

CD4-expressing T-cells, which recognize MHC Class II molecules, generally function as ‘helper’ cells: they help initiate immune responses including B-cell responses. CD4 cells fall into two main classes according to characteristic patterns of cytokine secretion. Activated ‘Th1’ cells release interferon-g and interleukin-2, and lie at the heart of generating target-specific inflammatory reactions,a classical example being the tuberculin skin test. Th2 CD4 cells similarly secrete interleukin-2, but also interleukins 4, 5, 6, 10, and 13, and transforming growth factor β. They are vitally important in B-cell stimulation and maturation, and also act antagonistically to many Th1 cytokines and so help regulate the activity of Th1 cells. CD8 lymphocytes bind to major histocompatibility complex Class I molecules; they are often but not exclusively cytotoxic in activity.

The initial characterization of CD4 and CD8 as ‘helper’ and ‘suppressor/cytotoxic’ cells, whilst helpful to a degree, is simplistic to the point of inaccuracy. Precisely the same may be said of the Th1–Th2 division; there is more likely a spectrum of cytokine secretion, and these represent two extremes. It may here also be re-emphasized that antigen presentation by antigen presenting cells to T-cells is a prerequisite for the development of the great majority of B-cell antibody, as well as cellular immune reactions, stressing the fundamental importance of T-cells in the generation of immune responses. T-lymphocytes indeed play a key role in initiating and coordinating virtually every limb of the immune system, both through direct contact with other cells and by the secretion of numerous lymphokines, including interleukins and interferon gamma. Thus the fundamental distinction of ‘T-cell’ responses versus ‘B-cell’, or ‘cellular’ and ‘humoral’ immune reactions, is likewise a serious over-simplification.

While the conventional dogma that the central nervous system is not routinely patrolled by lymphocytes is no longer tenable, any small volume of T-cell traffic would be unlikely to generate an immune response unless antigen-presenting cells were encountered, and also blood–brain barrier impairment seen, allowing other immune and inflammatory mediators to enter the central nervous system (Owens et al. 2002). The distribution of Class II major histocompatibility complex products gives some indication of individual cellular potential for antigen presentation and so may help identify the initial site of T-cell activation; and it is now clear that, contrary to initial reports, HLA antigens are expressed in the central nervous system.

In the normal central nervous system, microglia are the major cell population constitutively expressing Class II major histocompatibility complex molecules; this is not surprising since this population represents resident cells in the central nervous system of the macrophage-monocyte lineage, and is ultimately derived from the bone marrow. Interaction of T-cells with microglia, triggering secretion of cytokines, amplifies microglial Class II expression and acts on local endothelia, impairing blood–brain barrier function and recruiting further inflammatory and immunologically active cells.

In vitro, cytokines also induce the expression of Class I and II major histocompatibility complex products on astrocytes and cerebral vascular endothelial cells, both of which can present antigen to T-cells. Neither cell type appears normally to express Class II products in vivo, but cytokines such as g-interferon generated by T-cell/microglia interactions could induce Class II expression. Astrocytes can express intercellular adhesion molecule, ICAM, and may also secrete interleukin-1. Astrocytes and endothelial cells could at least in theory therefore have a central role in amplifying T-cell reactions and, since both are involved in maintaining normal blood–brain barrier physiology, in augmenting blood–brain barrier damage. It must be emphasized, however, that most results supporting this role derive from in vitro studies, with rather sparse supportive in vivo evidence.

In generating through stochastic recombination events to form T-cell receptors, the necessarily enormous diversity of epitope recognition required for the effective recognition of foreign antigens, chance dictates that many such receptors potentially recognize self antigens: against which an immune response would of course be detrimental, the underlying mechanism indeed of auto-immune diseases (Anderson et al. 2001). To help prevent such unhelpful behaviour, a number of mechanisms have evolved to allow ‘tolerance’, or active immune neglect, of self antigens. First, well over 90 per cent of T-cell antigen receptors generated in the thymus during T-cell genesis are clonally deleted, killed usually by apoptosis, and so never leave the thymus. Clonal deletion also occurs peripherally when, for example, very high levels of antigen are presented to the specific T-cell. The complex dance of antigen processing and presentation to the T-cell antigen receptor within a highly specific context needed for generating specific immune reactions also helps prevent inadvertent responsiveness to normally present self molecules.

Immune ‘networks’ also play a vital role in regulating the immune response. An immune response against a particular target triggers a secondary immune response directed against the very components of the primary response. This immune reactivity against specific antibodies and/or T-cell antigen receptors helps to suppress immunity in a highly target-selective manner. These are the anti-idiotypic responses and, complexly, in fact can in some circumstances stimulate the primary response too.

Steroids are surely amongst the commonest drugs prescribed in neurological practice. Azathioprine and methotrexate are used with increasing enthusiasm in neurology, but almost always for unlicensed indications. Intravenous immunoglobulins in the past few years have come to consume up to 50 per cent of neurology unit drug budgets, though in all likelihood this is currently being overtaken by the no less expensive interferons and glatirimer being used in significantly increasing numbers of multiple sclerosis patients. On this varied background, brief comment on some of these agents is required.

Glucocorticoids are notorious for their risks and side effects: these include a predisposition to infections, hyperglycaemia, hypertension, obesity, glaucoma and cataract, delayed wound and fracture healing, osteoporosis, psychosis, peptic ulceration and erosions, intracranial hypertension, avascular necrosis and steroid myopathy, together with adrenal suppression. When used in the longer term, alternate day regimens are generally thought to help reduce side effects although the evidence for this is not strong.

Nonetheless, their beneficial effects of course help justify a widespread use. Corticosteroids reduce oedema acutely, and suppress inflammation subacutely. They act by entering cells, binding to cytoplasmic receptors, and forming complexes which bind to steroid-responsive sites in nuclear DNA, thus influencing the expression of various species of mRNA. Corticosteroids produce multiple effects on the immune system although it is believed that they spare B-lymphocytes in vivo.

Methotrexate is a dihydrofolate reductase inhibitor, administered as a rule orally, once weekly.

Azathioprine is also an antimetabolite, likewise used with the aim of interfering with normal immune cell function, particularly proliferation, and both are commonly used in the treatment of cancer. Azathioprine acts by inhibiting purine biosynthesis and hence decreases the rate of cell replication particularly among B- and T-lymphocytes, resulting in both T- and B-lymphocytopoenia. Azathioprine takes up to 6 months to exert its full therapeutic effect; it is very poorly tolerated in up to 50 per cent of patients, mostly because of upper gastrointestinal symptoms, but prospective enzymatic testing of thiopurine methyltransferase, TPMT, can help predict those who will tolerate azathioprine and is recommended by some authorities, albeit without Class IV evidence and some studies cast doubt on its value (Kader et al. 2000; Sayani et al. 2005)).

Azathioprine has been associated with a risk of cancer, particularly lymphoma, and of infections through leucopoenia, and with hepatitis and pneumonitis. In relation to perhaps the most serious of these, careful studies have in fact failed to show a significant increased risk of cancer with medium-term duration of use of azathioprine in multiple sclerosis for 3–5 years or less, though 10 years’ adminstration or more may carry a more signficant risk (Confavreux, 1996; Taylor, 2004). Methotrexate carries a particular risk of hepatic fibrosis. Both drugs are also teratogenic, although careful analysis of published data suggest that evidence for this risk in patients, rather than experimental animals is poor, and the risk in fact very slight.

Mycophenolate is a more recently introduced anti-metabolite; it may be a more powerful immunosuppressant than azathioprine but carries a higher risk of malignancy.

Cyclophosphamide is yet more potent, acting as an alkylating agent, transferring alkyl groups to proteins, DNA, and RNA. It is thought particularly to affect B-cell proliferation, and so function. Bone marrow suppression, increased malignancy (especially urothelial), and haemorrhagic cystitis, teratogenicity, and sterility are amongst its more severe side effects (Section 36.2.6).

Agents such as tacrolimus, or FK506, cyclosporine, and rapamycin or sirolimus are relatively more recently introduced agents. Though structurally varied, they share the action of potently inhibiting calcineurin in the presence of their respective common ligands: the cytoplasmic immunophilins, cyclophilin, and FK506-binding protein. Immunophilins are in fact expressed in greater quantity in the central nervous system than in the immune system, and these drugs may well have useful neuroprotective effects in addition to their immune properties (Poulter et al. 2004) which are thought to lie more in suppressing T-cell function than B-cell. They are however significantly nephrotoxic, tacrolimus possibly more so than cyclosporin. Gastrointestinal disturbance, hepatic dysfunction, hypertension, arrhythmias, cardiomyopathy, psychosis, and encephalopathy also feature in their recognized side effects.

Mitoxantrone is a derivative of anthracyclin antibiotics, and acts as a cytotoxic agent. It exhibits dose-related cardiotoxicity and also carries a significant risk of malignancy, particularly drug-related acute myelogenous leukaemia of between 0.2 and 1 per cent. Nausea and alopecia also occur.

Immunomodulation A number of agents may be used to influence the activity of the immune system rather than simply suppress it. Immunoglobulins given intravenously have a variety of actions, interfering with anti-idiotypic networks, complement, and the activation and function of both T-cells, macrophages and microglia (Misra et al. 2005). Anaphylaxis or renal failure rarely occur as side effects; acute fulminant serum sickness is more likely in those with IgA deficiency.

Plasmapheresis superficially appears the obverse of intravenous imunglobulin therapy, its aim being to remove ciruclating immunoglobulins and also cytokines. The physical adventures of large vessel cannulation account for most potential hazards, but impaired electrolyte balance and haemostasis can also occur.

Interferon-b is a naturally occurring type 1 interferon made by many cell types, including dendritic cells and monocytes. It has a large number of effects on immune-active cells, on major histocompatibility complex Class I, with increases, and II, with decreases in expression, and on the blood–brain barrier. Precisely which effect(s) mediate its effect on relapses in multiple sclerosis is unknown (Billiau et al. 2004). Various side effects are reported, including commonly flu-like symptoms lasting from 3–4 to 24 h after delivery, injection site reactions, liver enzyme disturbances, and leucopoenias. Severe hepatitis is rarely reported, and depression may be precipitated or exacerbated.

Glatiramer acetate is also widely used in multiple sclerosis. It is a random copolymer of four amino acids in proportions resembling those found in myelin basic protein. Again, diverse actions are reported (Arnon et al. 2004), including the induction of glatiramer-specific T-cells and, interestingly, the stimulation of neurotrophin production by lymphocytes, but the mechanisms involved in the effects in multiple sclerosis are unknown. Urticarial and allergic reactions can occur but severe side effects are very uncommon.

This condition was first identified by Brain and co-authors in 1966. It presents variably either with episodic features: stroke-like events, relapsing encephalopathy, focal or generalized seizures, and/or psychotic episodes, or with a more insidious progressive pattern, with dementia, myoclonus in 50 per cent of cases, and also tremor. All occur in conjunction with high titres of anti-thyroid antibodies, usually anti-microsomal (Kothbauer et al. 1996).

The latter plainly are fundamental to the diagnosis, but their role in pathogenesis has yet to be established. It is possible, but thus far speculative, that anti-thyroid antibodies cross-react with brain antigens. Intrathecal synthesis of anti-thyroid antibodies, and the presence in CSF of immune complexes, has been demonstrated (Shaw et al. 1991; Ferracci et al. 2003). An alternative suggestion is that antibodies may, through immune complex formation, drive a putative vasculitic process; an association of Hashimoto’s thyroiditis with vasculitis in other tissues has been reported, specifically, with giant cell arteritis and vasculitic peripheral neuropathy. However, post-mortem histopathological investigations have shown little change or only mild, diffuse perivascular lymphocytic infiltrates (Brain et al. 1966).

Hashimoto’s encephalopathy exhibits female:male ratio of up to 9:1. Most cases are clinically and biochemically euthyroid at presentation. Imaging by CT or even MR is often normal, as is angiography, though isotope brain scanning may show patchy uptake. Very high titres of anti-thyroid antibodies are found, usually anti-microsomal. Spinal fluid examination may reveal a raised protein level and/or a raised cell count in 80 per cent of patients.

Most patients respond very well to steroid treatment; some have received further immunosuppressive therapy, such as cyclophosphamide, azathioprine, or plasmapheresis.

Dysthyroid eye disease is likely to be immunologically driven. Circulating Thyroid stimulation hormone receptor-stimulating antibodies cross-reactive with orbital fibroblasts are found. Thyroid-stimulating antibodies from patients with Graves’ ophthalmopathy can also stimulate fibroblast collagen synthesis. The orbit and extraocular muscles are oedematous and infiltrated with inflammatory cells and glycosaminoglycans, resulting in proptosis and a restrictive ophthalmopathy (Section 13.4.2). Upgaze limitation is the commonest presenting sign, though other muscles are also involved and ocular irritation is frequent. Vision is occasionally threatened by a complicating infiltrative or compressive optic neuropathy. Steroid treatment and radiotherapy appear to be equally effective (Bartalena et al. 2005).

Coeliac disease, or non-tropical sprue, is an immunologically mediated disorder resulting from intolerance to dietary gluten; it causes weight loss with steatorrhoea and/or diarrhoea, and malabsorption. In common with other enteropathies, neurological sequelae of a predictable nature may complicate coeliac disease as a direct consequence of malabsorption. Central nervous system complications apparently unrelated to deficiency states may also occur in perhaps 10 per cent of patients. Rarely, vasculitis is responsible, but the cause of the most commonly described and distinctive central nervous system association, a progressive cerebellar or spinocerebellar degeneration, with eye movement disorders, myoclonus, and occasionally epilepsy, remains unresolved. This was first described in 1966 (Cooke et al. 1966), but a provocative report 30 years later proposed that a substantial proportion of patients with idiopathic ataxia had this disorder, often in the absence of bowel symptoms (Hadjivassiliou et al. 1998). Others, however, suggested the association is rare and may not be causal (Lock et al. 2005).

Stiff Man, or stiff-person, syndrome (Sections 23.7.2; 40.10.3) is an uncommon disorder (Moersch et al. 1956; Dalakas et al. 2000) generally now agreed to be of autoimmune origin. It is associated with anti-GAD, glutamic acid decarboxylase antibodies which may, it is thought, cross-react with and affect a specific sub-population of spinal neurones. It is associated with diabetes mellitus, or with systemic autoimmune diseases, particularly lupus, and can be seen as a paraneoplastic disorder, though in the majority of cases it occurs in isolation.

It presents with adult onset slowness, aching discomfort and stiffness of muscles, mainly but not exclusively, axial, and with painful muscle cramps, progressing slowly over months and years (Dalakas et al. 2000). Spasms, often noise-, startle-, or action-induced, may be very severe; tendon and muscle rupture may occur. Walking may become clumsy ion reveals normal power and tendon reflexes, downgoing plantar responses, and no abnormalities either of sensation or, barring spasms, coordination. However, axial and abdominal wall rigidity is apparent, and there may be proximal limb muscle stiffness, agonists and antagonists acting simultaneously. A hysterical origin for the symptoms is often wrongly assumed. Asymmetrical contraction of the paraspinal muscles causes a characteristic lordotic and often scoliotic posture.

Brain and spinal cord imaging is normal. The spinal fluid is usually normal but for the common finding of oligoclonal immunoglobulin bands. Electrophysiological muscle examination reveals continuous muscle activity despite invitation to relax, with normal motor unit morphology; ‘The patient was unable to relax during the examination’ should raise suspicion. Importantly, voluntary contraction of antagonists fails to inhibit the activity in the muscle under examination. Abnormal activity, and likewise spasms, does not persist during sleep; its central origin is confirmed by its disappearance following pharmacological peripheral nerve block or spinal or general anaesthesia, in contrast to the abnormal activity demonstrable in neuromyotonic syndromes.

The syndrome is thought to result from an imbalance between excitatory catecholaminergic and descending inhibitory g-amino butyric acid or GABA-ergic influences on spinal motor neurones. The finding of increased brainstem excitability is consistent with a widespread dysfunction of central inhibitory mechanisms (Molloy et al. 2002). Antibodies directed against glutamic acid decarboxylase, the enzyme responsible for producing GABA from glutamic acid, which therefore react with GABA-ergic neurones, and also with pancreatic islet β-cells, are present in 60 per cent of patients. A clonal B-cell response against glutamic acid decarboxylase is apparent within the CSF, partly accounting for the oligoclonal immunoglobulin bands (Dalakas et al. 2000).

Interestingly, anti-GAD antibodies are reported now also to occur in patients with Batten’s disease (Section 10.3.2). Interpretations of this observation could not be more varied: either their occurrence in a hereditary neurodegenerative condition suggests they can arise as an epiphenomenon, or their presence suggests an additional and important immune component to the pathogenesis of Batten’s disease (Dalakas 2005). The two possibilities are not, of course, mutually exclusive.

Benzodiazepines particularly, tizanidine, and also baclofen, and occasionally sodium valproate are used therapeutically. More experimental treatments have included intrathecal baclofen and paraspinal botulinum toxin. There is now Class 1b evidence for the value of intravenous immunoglobulin (Dalakas et al. 2001; Dalakas 2005).

In patients with cancer and Stiff Man syndrome, anti-neuronal antibodies of a different specificity, to a synaptic-vesicle-associated protein amphiphysin, may be found.

A more serious condition is progressive encephalomyelitis with rigidity, where stiffness is accompanied by cranial neuropathies, myoclonus, ataxia, diminished tendon jerks, and extensor plantar responses, MRI brainstem and spinal cord changes occur, and the CSF shows a pleomorphic leucocytosis. The course is substantially more aggressive, with death in 3–10 years. It is often paraneoplastic, and anti-GAD or anti-amphiphysin antibodies can be found.

The term ‘channelopathy’ first entered the neurological lexicography in relation to inherited disease: mutations in genes encoding voltage-gated calcium, chloride, sodium and potassium channels, and also of various ligand-gated channels, such as glycine, GABA, and nicotinic acetylcholine receptors all have been implicated in central nervous system and peripheral nervous system diseases ranging from epilepsy and ataxia to myotonia.

Subsequently, disorders arising as a consequence of antibodies directed against these ion channels came to be included in the new order as acquired channelopathies, although a detailed understanding of the pathophysiology of the archetypal disorder myasthenia gravis long preceded this terminology. Acquired neuromyotonia, Isaac’s syndrome, related to antibodies to voltage-gated potassium channels, is also discussed elsewhere (Section 23.7.1).

A third neuromuscular disorder, the Lambert–Eaton myasthenic syndrome, is related to antibodies directed against voltage-gated P/Q-type calcium channel and is also more comprehensively described elsewhere (Section 24.10.2). Approximately two-thirds of patients exhibit Lambert–Eaton syndome as a paraneoplastic pheno-menon, the remaining third appearing to have a primary autoimmune antibody disorder. Up to 10 per cent of patients also have cerebellar ataxia. The same voltage-gated P/Q-type calcium channel is expressed by cerebellar Purkinje and granule cells, and such patients show Purkinje cell loss at post-mortem.

Another mix of a paraneoplastic central nervous system phenotype and neuromuscular channelopathy is Morvan’s syndrome, an association of limbic encephalitis with neuromyotonia plus peripheral neuropathy and hyperhidrosis (Liguori et al. 2001). Here, antibodies to voltage-gated potassium channels, VGKCs, have been found in the sera and cerebrospinal fluid. Plasma exchange reduces the concentration of serum voltage-gated potassium channels antibodies and this is associated with clinical improvement. A significant proportion of such patients appears to have no tumour ‘driving’ antibody production. Antibodies to voltage-gated potassium channels have been increasingly associated with acquired acute or subacute epileptic and encephalopathic clinical presentations, particularly of an amnesic or limbic type (Vincent et al. 2004; McKnight et al. 2005), responsive to plasma exchange. These are in the main autoimmune, non-paraneoplastic, and potentially treatable.

Paraneoplastic syndromes are a group of disorders caused by a malignancy, but not occurring as a direct structural consequence of the cancer or any metastases. The term is often used synonymously with the remote effects of cancer, and does not usually include more general non-metastatic manifestations of malignancy, such as fever, malaise, and lethargy, nutritional disorders, infection, and iatrogenic disorders.

Paraneoplastic syndromes may involve many organs apart from the nervous system:

Neurological paraneoplasia is not common, occurring perhaps in less than 1 per cent of patients with cancer, but may cause profound morbidity. Almost any part of the central or peripheral nervous system may be involved, and a number of classical and stereotyped syndromes have been described. Each may be associated with characteristic serum antibody (Table 38.1), and the combination of antibody plus clinical syndrome essentially secures the diagnosis. From the diagnostic perspective, however, it is important to note that often there is much overlap both in clinical syndrome and antibody association. Patients with small cell lung carcinoma, for example, often harbour several autoantibody types (Pittock et al. 2004).

Finally by way of introduction, these syndromes may precede symptoms more directly resulting from the tumour by months, or rarely years. Indeed often the cancer is identified only at post-mortem examination.

Paraneoplastic cerebellar degeneration may occur in various contexts of malignancy (Brain et al. 1965b). Many patients have cervical, uterine, ovarian, or breast cancer: they may harbour anti-Yo antibodies (Peterson et al. 1992). Hodgkin’s disease is another precipitant, this is more common in men, and associated with anti-Tr antibodies (Trotter et al. 1976). Small cell lung cancer can also precipitate this disorder, often with no identifiable antibody, or occasionally with anti-Hu antibodies. In less than 50 per cent of cases, a prior diagnosis of cancer has been made.

The commonest presentation is with a subacutely progressive ataxic syndrome, usually in late middle age or above, reflecting the age range of the underlying cause. What has been termed a vermis phenotype is perhaps the most typical, with prominent gait and truncal ataxia and relative sparing of the limbs; disability is profound. It has been commonly observed that in many patients the syndrome may progress to a plateau and thereafter remain stable for considerable periods, irrespective of the progression or otherwise of the underlying tumour. Nystagmus, often downbeat, is usually present. Other neurological features may be present, including sensory symptoms, dysphagia, and extensor plantar responses. MRI can demonstrate cerebellar atrophy, but is more commonly normal at presentation.

Spinal fluid examination may reveal an elevated protein level; there may be a modestly raised lymphocyte count. The occasional presence of oligoclonal immunoglobulin bands completes a CSF pattern which is common to each of the paraneoplastic neurological syndromes.

A number of clinical phenotypes, often occurring in admixture with one another, and sharing a common underlying malignancy, which is small cell lung cancer, in 80 per cent cases, neuropathology, and frequent antibody association, anti-Hu, fall within the spectrum of encephalomyelitis:

One of the commoner paraneoplastic manifestations (Denny-Brown 1948; Croft et al. 1965; Hughes et al. 1996), this syndrome is often grouped with the preceding encephalomyelitides, again sharing small cell lung carcinoma as the commonest precipitant and a common anti-Hu antibody association; indeed in up to 50 per cent of cases, it coexists with encephalitic features (Chalk et al. 1992). Autonomic failure may be prominent (Siemson et al. 1963). Approximately 1 per cent of patients with small cell cancers develop clinical or electrophysiological evidence of sensory neuronopathy (Elrington et al. 1991), but the disorder is also seen in breast and other cancers (Horwich et al. 1977; Hughes et al. 1996).

The neuronopathy is characteristically painful from the onset and accompanied by often distressing parasthaesia (Section 21.13.1). Symptoms commence in the extremities and progress, proximally sometimes very rapidly, and examination reveals the not unexpected findings of distal sensory loss to all modalities, including vibration and joint position sense, with areflexia and often sensory ataxia. Cranial sensory nerves may be involved. Electrophysiological testing shows absent sensory action potentials and no motor disturbance (Donofrio et al. 1989). The disorder is so-named, rather than as sensory neuropathy, on histopathological grounds, the dorsal root ganglia showing changes, not peripheral nerve. A clinically identical syndrome may occur in association with Sjögren’s syndrome (Font et al. 1990).

The occurrence of a syndrome resembling amyotrophic lateral sclerosis phenotype, with mixed upper and lower motor neurone signs, in the context of cancer, has been intermittently reported since the first description four decades ago (Brain et al. 1965a). It has, however, remained controversial: the often quoted figure of up to 10 per cent of cases of amyotrophic lateral sclerosis associated with malignancy (Norris et al. 1965) certainly falls well outside most neurologists’ experience.

A subacute motor neuronopathy without pyramidal signs, associated especially with lymphomatous malignancies is a more secure entity (Schold et al. 1979). The syndrome is progressive and painless, and bears some clinical resemblance to multifocal motor neuropathy with conduction block, though the latter most characteristically affects the upper limbs more than the lower; paraneoplastic motor disease mainly the reverse. The distinction is more reliable if made electrophysiologically, the self-evident features of the former contrasting with the normal motor and sensory conduction of paraneoplastic motor neuronopathy accompanied by electromyographic evidence of denervation. Neither upper motor neurones nor the bulbar lower motor neurones are usually affected. Nuchal and respiratory muscle weakness may occur; anti-Hu antibodies and subclinical loss of cerebellar Purkinje cells are reported (Verma et al. 1996).

Such a motor neuronopathy occurring with a limited myelitis might generate the clinical picture of amyotrophic lateral sclerosis. It has been suggested that the eleven cases of cancer-related motor neurone disease described by Brain in 1965 may have represented ‘burnt out paraneoplastic encephalomyelitis’ (Posner 1996). However, Younger et al. (1991) reported nine patients, all with lymphoma, eight of whom had clinical features suggesting amyotrophic lateral sclerosis (Younger et al. 1991). Others have reported series of patients with lymphoproliferative disorders and motor neurone disease with no motor neuropathy, most having ‘definite or probable’ upper motor neurone signs (Gordon et al. 1997). Of these only the small minority with purely lower motor neurone disease showed a neurological response to treatment for their lymphoproliferative disease. An association in very small numbers of patients of motor neurone disease with anti-Hu antibodies has also been reported, with a possible link with breast cancer (Forsyth et al. 1997).

Breast cancer may also be associated with two relatively discrete disorders, a sensorimotor neuropathy (Peterson et al. 1994), and the so-called ‘numb chin syndrome’, characterized by numbness in the distribution of the mental or alveolar branches of the mandibular nerve (Horton et al. 1973; Burt et al. 1992; Lossos et al. 1992). Other malignancies may precipitate the latter, which is thought to represent a direct effect of metastasis, rather than paraneoplasia.

This is a profoundly disabling, distressing syndrome, wherein opsoclonic eye movements cause profound vertigo, nausea and anorexia, and debility, usually confining the patient to bed (Section 13.2.5). The eye movement disorder consists of chaotic, involuntary partial or complete saccades which are arrhythmic, continuous, and randomly directed. The movements persist during sleep. Truncal and limb myoclonus may also be present, as may central ataxia (Anderson et al. 1988a). Some such patients have anti-Ri antibodies and breast or gynaecological cancers (Luque et al. 1991), but many harbour cancers and no obvious antibody (Bataller et al. 2001). In adults, small cell lung cancer is yet again the commonest associated malignancy, and a tendency to remit is described (Anderson et al. 1988a).

The syndrome was first described not in adults but in children with neuroblastoma (Kinsbourne 1962), the underlying cause in approximately 50 per cent of children with this neurological picture (Telander et al. 1989). It occurs too as a monophasic post-infectious process in adults, without cancer, usually following respiratory or gastrointestinal infection (Baringer et al. 1968). In one study of 58 patients, only 11 had identified cancers.

This syndrome again is most commonly associated with small cell lung carcinoma, though other tumours may also be responsible:gynaecological malignancy and, perhaps particularly, melanoma (Section 12.3.2). Antibodies directed against the calcium-binding protein recoverin are found in some patients with paraneoplastic retinal degeneration (Polans et al. 1991; Thirkill et al. 1992). Paraneoplastic retinal degeneration precipitated by melanoma is clinically distinct (Kim et al. 1994). Serum antibodies are associated; they are directed not against recoverin but against bipolar retinal neurones (Weinstein et al. 1994). Antibodies directed against optic nerve cells are also described, and serum antibodies directed against enolase have been demonstrated in patients with retinopathy associated with a variety of malignancies (Adamus et al. 1996).

Photosensitivity, night blindness, visual scotomata, and retinal artery attenuation are typical features, though variations on this phenotype are well-described (Jacobson et al. 1990). Visual loss is painless, and may be monocular initially. Visual evoked responses usually reveal normal optic nerve function, electroretinography confirming the site of derangement. A paraneoplastic optic neuropathy, associated with anti-CV-2 antibodies, is also recognized.

The association of inflammatory muscle disease with malignancy has for some years been considered overstated, but most now accept that dermatomyositis in particular is significantly linked with cancer (Sigurgeirsson 1992; Hill et al. 2001), though polymyositis only weakly. The clinical features and histopathology are not distinguishable from those of idiopathic polymyositis (Section 24.7.2).

The principal pathological picture associated with the described syndromes is strikingly similar, though with a few notable exceptions, across the spectrum of paraneoplastic conditions (Denny-Brown 1948). The main features may be divided into three groups:

Thus in paraneoplastic cerebellar degeneration these changes occur in the cerebellar cortex (Brain et al. 1965b), where Purkinje cells are selectively lost. In encephalomyelitic syndromes, striking changes are found, according to the specific disorder, respectively in the limbic system (Corsellis et al. 1968), the brainstem (Henson et al. 1965), or the spinal cord (Mancall et al. 1964), while in subacute sensory neuronopathy, the pathology is centred upon the dorsal root ganglion (Denny-Brown 1948; Croft et al. 1967; Horwich et al. 1977). In later stages, the changes may extend proximally into the posterior columns, and distally into the peripheral nerve; demyelination is a marked feature in a minority of patients. In subacute motor neuronopathy, anterior horns in the spinal cord bear the brunt of the disease process; in contrast with amyotrophic lateral sclerosis, the lateral columns are not affected (Brain et al. 1965b; Henson et al. 1965). The findings in cancer-associated retinopathy are again typical, with loss of photoreceptors and retinal ganglion cells accompanying the inflammatory changes (Grunwald et al. 1987). It should also be noted that in most of these syndromes, less pronounced findings of a similar nature may be present diffusely in the cerebral hemispheres, brainstem, and spinal cord.

In paraneoplastic opsoclonus–myoclonus syndrome, the picture is a little less clear. Eye movement physiology might volunteer ‘omnipause cells’ as the target cell; these are inhibitory interneurones putatively located in the pontine reticular formation. In some patients, however, no clear abnormalities are identifiable here or indeed elsewhere (Ridley et al. 1987; Anderson et al. 1988a); in others, the findings are identical to those of paraneoplastic cerebellar degeneration (Ellenberger et al. 1968).

The most economical mechanistic interpretation of these changes would be that an immunological or inflammatory process is targeted upon certain neuronal sub-populations. The classical and obvious hypothesis has been that specific antibodies directed against tumour surface antigens arise as part of the anti-tumour immune response. Certain neuronal antigens may be expressed by tumour cells, and tumour-specific antibodies directed at these targets cross-react with and damage certain neuronal populations. Thus, small cell lung cancer cells may express P/Q type voltage-gated calcium channels, antibodies to which are considered responsible for causing Lambert–Eaton myasthenic syndrome; acquired channelopathy here meeting paraneoplasia.

Central nervous system paraneoplasia is less straightforward, however. Anti-Yo antibodies, for example, found in many patients with paraneoplastic cerebellar degeneration, and almost invariably in those with an underlying breast or gynaecological malignancy rather than those with small cell lung cancer, do indeed react with Purkinje cell antigens. They are also expressed in the tumour. However, the antigens are cytoplasmic, not surface-expressed; CD62 is a DNA-binding protein which directs gene transcription, of the leucine-zipper family (Sakai et al. 1990; Fathallah et al. 1991). How or indeed if these antibodies interact with these cryptic antigens in vivo, and whether this interaction is responsible for Purkinje cell loss, is unclear, but anti-Yo antibodies are not detected in non-paraneoplastic cerebellar degenerations (Anderson et al. 1988b; Smith et al. 1988; Peterson et al. 1992), arguing against secondary generation triggered by other causes of Purkinje cell damage.

Purkinje cells exposed to anti-Yo/anti-Purkinje cell antibodies antibodies can specifically take up these immunoglobulin molecules (Graus et al. 1991; Greenlee et al. 1995) but, perplexingly, no cell injury is apparent in these cell culture models. There is little or no evidence from other immunological systems that internalized antibody can cause cell damage (Naparstek et al. 1993). Active immunization with recombinant Yo protein also fails to induce cerebellar disease (Tanaka et al. 1994; Sakai et al. 1995) while passive transfer of Yo/APCA-specific lymphocytes into immune-deficient mice similarly fails to replicate disease (Tanaka et al. 1995).

Likewise, anti-Hu-related antigens are also expressed in small cell lung cancer cells (King et al. 1996), and in neurons, but here ubiquitously by cells in the brain and spinal cord, and also in dorsal root ganglian sensory neurones and in autonomic ganglia (Dick et al. 1988; Altermatt et al. 1991). As with anti-Yo antibodies, intrathecal immunoglobulins of these specificities are found in patients with paraneoplasia. HuC, HuD, Hel-N1, and Hel-N2 share sequence homology with each other and appear to be RNA-binding proteins important in post-transcriptional gene processing (Dropcho et al. 1994). Again they are not expressed on the cell surface, but in the nucleus, and attempts to develop animal models of anti-Hu/ANNA-1-related disease by injecting antibody have repeatedly failed (Dick et al. 1988).

Anti-Ri antibodies react with breast tumour tissue from patients with opsoclonus, and are non-reactive with malignant breast tissue not associated with opsoclonus (Luque et al. 1991). They react with nuclear, not surface antigens, namely the RNA-binding protein Nova (Buckanovich et al. 1993).

However, in relation to the cancer-associated-retinopathy-associated antigen recoverin, although intracellularly located within photoreceptor cells in the retina, a different story is emerging. Recoverin expression in tumour tissue from a patient with cancer-associated retinopathy is found (Polans et al. 1995), and the presence of retinal deposits of immunoglobulin, with loss of ganglion cells, in immunohistopathological studies of patients (Grunwald et al. 1987) provides further evidence for an autoimmune basis to cancer-associated retinopathy. In marked contrast to the various paraneoplastic antibodies and antigens described above, an animal model of cancer-associated retinopathy has been described: Lewis rats injected with a synthetic peptide fragment of recoverin develop photoreceptor cell degeneration (Polans et al. 1995). Furthermore, anti-recoverin antibodies cause apoptotic death of retinal cells in vivo (Adamus et al. 1997). Whether this offers mechanistic clues to other paraneoplastic disorders remains to be established. Alternatively, cell surface expression of anti-Hu-related antigens by small cell lung cancer cells has been reported (Tora et al. 1997), with obvious and important implications for pathogenesis.

Other possibilities are, however, considered. In paraneoplastic motor neurone disease, a possible opportunistic infectious aetiology has been suggested. The particular association with lymphoma offers immune paresis as a contribution to this suggestion; poliomyelitis represents an obvious example of viral disease of the anterior horn cell.

Paraneoplastic anti-neuronal antibodies have assumed an important diagnostic role of great practical help in investigating patients. The difficulty in establishing a pathogenetic contribution has encouraged studies of T-cell involvement (Albert et al. 2000). In addition to B-cells, macrophages and T-lymphocytes are prominent in the lesions of paraneoplastic encephalomyelitis. CD4- and CD8-positive cells are present.

The diagnosis is made on the basis of these characteristic clinical phenotypes arising in conjunction with pertinent antibodies found on serological testing. Some syndromes, for example, opsoclonus–myoclonus in an adult, are sufficiently suggestive of paraneoplasia to warrant extensive searching for malignancy even in the absence of antibodies. In most cases, whether with or without antibodies, the underlying malignancy will either already be known, or be disclosed by conventional detailed physical examination and ‘first line’ tests. Should this not be the case, FDG-positron emission tomography scanning now has a recognized role in revealing cryptic cancers (Rees et al. 2001).

The most obvious therapeutic approach is to treat the causative underlying malignancy (Batson et al. 1992): cure of the cancer should, surely, cure the neurology given time? Anecdotal reports attest to occasional striking neurological responses. The author has witnessed a severely amnesic patient with small cell lung cancer and limbic encephalitis improve in a clear, substantial, and carefully documented way following treatment of the lung tumour. Of course, when the tumour has generated neuromuscular disease, the greater capacity for repair of the affected neurological target allows tumour removal to be a successful means of treating the syndrome, for example, thymectomy in myasthenia, and lung tumour treatment in Lambert–Eaton myastheric syndrome.

For such to be the case consistently, however, one would have to postulate temporary mechanisms of neuronal dysfunction and not cell death. For syndromes persisting for many months this may be unlikely at least in central nervous system disease. The common clinical course of most paraneoplastic disorders of reaching a plateau after a period of subacute progression most likely suggests the acquisition of irreversible neuronal damage (Dropcho 1995). This may, help to explain ‘stabilization’ as an apparent partial response to well-timed interventions. Not surprisingly therefore, and despite these occasional reports of responding patients (Paone et al. 1980), when relatively large series of patients are studied, little objective support is seen for tumour removal as neurological therapy (Graus et al. 1995; Grisold et al. 1995).

Interestingly, the presence of anti-Hu antibodies in small cell lung cancer patients without paraneoplastic neurological disease appears to correlate with a better overall response to cancer treatment, consistent with an anti-tumoural protective role for the antibody (Graus et al. 1997).

This finding emphasizes the potential disadvantage of the alternative approach to treating neurological symptoms. This consists of immunotherapies directed not against the tumour, but towards the immune mediators putatively responsible for the neurological syndrome: that is, the antibodies themselves. Again considering for example the well-documented neuromuscular paraneoplastic disorders myasthenia gravis and Lambert–Eaton myasthenic syndrome, the clear response to plasmapheresis provides some support for this approach, though of course only as symptomatic therapy (Newsom Davis and Murray 1984). Of course, the incurability of many of the underlying malignancies in central nervous system paraneoplasia, combined with the extreme and unremitting distress, discomfort, and disability caused by many such disorders, and the absence of any other useful treatments, even palliative, must leave no possible therapeutic avenue unexplored.

Thus both conventional and novel immunotherapies have been exhibited. Steroids may help in paediatric opsoclonus–myoclonus (Boltshauser et al. 1979), and it has been suggested that adult opsoclonus–myoclonus may be a more benign disease, and more commonly steroid-responsive, than other paraneoplastic disorders (Dropcho et al. 1993). However, those with much experience have not found immunosuppression to be successful in other situations (Dalmau et al. 1996). Also the possible adverse effects of inadvertently suppressing anti-tumour immune surveillance and reactivity mitigate against non-specific immunosuppression.

Isolated case reports or small series attest to the possible merits of more immunoglobulin-specific approaches such as Intravenous immunoglobulin (Counsell et al. 1994) and plasma exchange (Weissman et al. 1989), but in larger series, no significant improvement emerged (Graus et al. 1995). In 18 patients suffering anti-Hu- related paraneoplastic disorders treatment with Intravenous immunoglobulin produced no benefit in seriously affected patients; (Vega et al. 1994), similarly, in a study of 22 patients with a variety of antibody-associated paraneoplastic disorders, no significant evidence of benefit was found (Uchuya et al. 1996); the same applies in respect of plasma exchange. Protein A immunoadsorptive columns have also been used in small numbers of patients (Cher et al. 1995; Nitschke et al. 1995).

The majority of patients with paraneoplastic diseases of the central nervous system do not recover neurologically; their disability may become static rather than continuing to progress, and their overall prognosis depends on that of the underlying malignancy. Non-immunological symptomatic treatment assumes great importance (Brady 1996).