Paraneoplastic Syndromes

The fact that a small cancer hidden in the chest, abdomen, or pelvis could destroy or damage portions of the nervous system, such as cerebellar Purkinje cells or cholinergic synapses, has intrigued neurologists since paraneoplastic syndromes were first described. In 1965, when little was known about their pathogenesis, a full issue of the journal Brainand an international symposium were devoted to paraneoplastic disorders. In this decade, the discovery of several paraneoplastic antibodies that react with both the nervous system and the causal cancer has rekindled interest in these syndromes (Table 1). Several other factors make these rare syndromes of clinical and scientific interest. A recent review by Dalmau and Posner¹contains a more comprehensive bibliography of paraneoplastic syndromes.

These disorders challenge the diagnostic skills of the neurologist. The patient's cancer has usually not yet been discovered and, because other inflammatory disorders of the nervous system can mimic paraneoplastic syndromes, the diagnosis is often difficult. However, paraneoplastic antibodies, in most (but not all) instances, unequivocally establish that the disorder is paraneoplastic.

The disorders challenge the diagnostic skills of the oncologist. The underlying cancer may be so small as to be undetectable by even the most sophisticated imaging techniques. Fortunately, many paraneoplastic antibodies also point to the most likely underlying tumor. For example, the anti-Hu antibody indicates the presence of a small cell lung cancer and the anti-Yo antibody, the presence of an ovarian or breast cancer. So strong is the association of the anti-Yo antibody with gynecologic cancers that we and others recommend hysterectomy and salpingo-oophorectomy in anti-Yo–positive postmenopausal patients with normal mammograms, even in the absence of positive imaging studies of the pelvis. In only 2 patients of whom we are aware has a cancer not been found at pelvic surgery. In both patients, cancer of the breast was subsequently discovered. The anti-Ta antibody seems to be specific for testicular cancer.

Paraneoplastic syndromes are a therapeutic challenge for the neurologist. With the exception of myasthenia gravis, the Lambert-Eaton myasthenic syndrome, neuromyotonia, dermatomyositis, and certain peripheral neuropathies associated with myeloma, treatment of paraneoplastic syndromes is generally unsatisfactory.

Paraneoplastic syndromes are a therapeutic challenge for the oncologist. Substantial evidence suggests that in patients with paraneoplastic antibody-positive serology, the neoplasms grow more indolently and are less likely to metastasize than in patients with the same cancer who are not antibody positive or who do not have paraneoplastic symptoms.²Recent animal data support these conclusions.³If the neurologist chooses to treat the neurologic symptoms with immune suppression, it is possible that the oncologist will be faced with a more rapidly growing tumor.

Paraneoplastic antibodies react with both the cancer and the nervous system. These antibodies identify antigens, present normally only in the nervous system (usually in neurons), but for uncertain reasons expressed ectopically in certain tumors. The immune system recognizes the proteins expressed by the tumor as foreign and mounts an immune attack that partially controls tumor growth (in some instances it appears to destroy the tumor so that no tumor is found even at autopsy). The immune reaction also attacks portions of the nervous system that express the antigen. The pathogenesis of some syndromes (myasthenia gravis, Lambert-Eaton myasthenic syndrome, neuromyotonia) is mediated by antibodies but, for the majority of antibody-associated paraneoplastic syndromes of the central nervous system (CNS), the major pathogenic mechanism appears to be related to cytotoxic T-cell responses.⁴

In paraneoplastic syndromes affecting the CNS, inflammatory infiltrates of T cells and plasma cells are found both in the nervous system and in the cancer. Furthermore, IgG normally absent from the CNS can be found not only within the neuropil but also within neurons. Elution studies indicate that the antibody found in the brain and the tumors of patients with paraneoplastic syndromes is the same antibody that reacts with onconeural antigens.⁵

Because onconeural antigens are normally present only in neurons, it is likely that they are important for neural development and phenotype maintenance. Accordingly, we and others have used paraneoplastic antibodies to probe complementary DNA expression libraries to clone the antigen(s) identified by these antibodies (Table 1).With respect to those antigens that have been identified so far, the molecular analysis appears to confirm that onconeural antigens are important to neuronal function. The remainder of this article will address the question of the function of some onconeural antigens.

The term Hu antigensrefers to a family of nuclear proteins normally expressed in all neurons of the central and peripheral nervous system but not in other cell types (with the possible exception of the testes). The antigen was probably first identified by Wilkinson and Zeromski⁶in 1965, when they reported that 4 patients suffering from subacute sensory neuronopathy associated with lung cancer had in their serum a low-titer antibody that reacted with the cytoplasm [sic] of neurons in the guinea pig cerebral cortex. No additional information was forthcoming until 1985, when Graus and colleagues⁷described first 2 and later 4 patients with subacute sensory neuropathy associated with small cell lung cancer who had in their serum high titers of a complement-fixing antibody that reacted predominantly with the nuclei of central and peripheral nervous system neurons. The Hu antigens correspond to a set of proteins of 35 to 40 kd on Western blot using either neuronal or small cell lung cancer protein extracts. Subsequently, it has become evident that the anti-Hu antibody is a marker not only of sensory neuronopathy but also of encephalomyelitis.

Subsequent work has identified several proteins belonging to the Hu family. All of these have high homology with the Drosophilaembryonic lethal abnormal visual (ELAV) protein, necessary for the development of the eye and the CNS of the fly. All Hu proteins, including HuD, HuC, Hel-N1, and HuR, contain 2 tandem RNA recognition motifs (RRM), a basic domain, and a third RRM. The basic domains of HuD, Hel-N1, and HuC are alternatively spliced yielding HuDpro, HuDmex, Hel-N2, and HuC isoforms. The expression of HuD, HuC, and Hel-N1 is restricted to neurons, but HuR is reported to be ubiquitously expressed in extraneuronal tissues.⁸These proteins bind to AU-rich elements, which are present in the 3′-untranslated region of messenger RNA (mRNA), that regulate cell proliferation (ie, c-fos, c-Myc, Gap43, and GM-CSF). The exact function of the Hu proteins remains unknown, but it has been postulated that they act as transfactors involved in selective mRNA degradation.

The anti-Hu antibodies bind to the first and second RRM of HuD, HuC, and Hel-N1 but not HuR; it is unknown how the binding of anti-Hu antibodies may affect the function of these proteins, but for HuD the first and second RRM are essential for RNA binding.

Although a study of the developing mouse nervous system has demonstrated a hierarchy of expression of the Hu proteins, in the adult mouse, all Hu proteins appear to be expressed in most brain regions.⁹The same study showed that dorsal root ganglia, which in about 70% of patients with the anti-Hu syndrome is the initial target of the disease, has a robust expression of the 4 mouse Hu homologs. The Hu proteins are an early marker of neuronal development.¹⁰Although not present in neuroblasts, they first appear in early lineage neurons that are still capable of proliferation. They appear to be the earliest unequivocal marker of neuronal commitment in CNS-stem cells. Hu-positive neurons that retain the capacity for proliferation can also be found in the lateral ventricle subependyma of adult rats¹⁰and in the temporal ventricular subependyma examined from autopsy of epileptic patients. The role of the Hu proteins in the development of neurons is indicated not only by their early expression in neurons still capable of proliferation but also by the fact that mutation of the Drosophilahomologous ELAVgene prevents development of the Drosophilanervous system.

HuD is expressed in a number of tumors, including all small cell lung cancers and mostly all neuroblastomas, as well as occasional other tumors (including several types of sarcoma and prostate carcinoma).

Whether, and if so how, the presence of antibodies against Hu proteins in the serum samples of patients with paraneoplastic syndromes is associated with the destruction of portions of the nervous system and with relative control of the growth of the underlying tumor is unclear. What role the Hu proteins play in small cell lung cancer and the other cancers in which they are expressed is also unclear. The HuD protein in small cell lung cancers from patients with or without paraneoplastic syndromes is identical to its neuronal counterpart and not mutated.

The term Yo proteinsrefers to a family of proteins highly expressed in the cytoplasm of cerebellar Purkinje cells and in the tumor cells (usually gynecologic or breast) of patients with anti-Yo–positive antibody paraneoplastic cerebellar degeneration (PCD). The anti-Yo antibody was first reported by Greenlee and Brashear¹¹in 1983, and later by Jaeckle et al¹²in patients with ovarian cancer and subacute cerebellar degeneration. In the rat or mouse brain, anti-Yo serum at low dilution reacts with cerebellar Purkinje cells as well as some other large cells in the CNS and, in at least 1 laboratory study, with Schwann cells. At higher dilutions, using human tissues, the reaction is restricted to Purkinje cells of the cerebellum, to the tumors that are associated with anti-Yo–positive PCD, and about 30% of ovarian cancers not associated with PCD. Surprisingly, mRNA for Yo proteins is widely distributed not only in the nervous system but throughout the body. The selective expression of Yo proteins by Purkinje cells appears to be posttranscriptionally regulated.

The anti-Yo serum identifies 2 bands in immunoblots of protein extracts of Purkinje cells. The bands had relative molecular weights of 62 and 34 kd. Subsequently, an intermediate band of about 52 kd has also been discovered.

There are 3 types of Yo proteins: CDR34, CDR62-1, and CDR62-2. The first, cloned and sequenced by Dropcho et al,¹³is a 34-kd protein that consists of 34 inexact tandem repeats of a hexapeptide. Amino acids 2, 3, 4, and 6 in this hexapeptide are invariably leucine, glutamate, aspartate, and aspartate, respectively. Amino acids 1 and 5 are more variable—1 being leucine or phenylalanine and 5 being methionine or valine. The function of this 34-kd protein is unknown. Anti-Yo antibodies recognize epitopes present in this protein.

Two 62-kd proteins (CDR62-1 and CDR62-2) have also been cloned and sequenced. These proteins have a leucine zipper motif, which contains the epitope(s) recognized by anti-Yo antibodies. The occurrence of leucine zipper motifs in the predicted open-reading frame of these proteins suggests that they may play a role in the regulation of gene expression. Analysis of tumors from patients with anti-Yo–associated PCD have demonstrated that the tumors express CDR34 and CDR62-1 but not CDR62-2. Recently, Darnell and colleagues (oral communication, 1999) have shown that the 62-kd protein binds to c-mycand have postulated that it exerts its activity by inhibiting the activity of the c-mycgene. The role of the anti-Yo antibody in causing PCD is unclear, but high titers of an antibody reacting predominantly with Purkinje cells in a disease characterized by loss of all Purkinje cells with relative sparing of the remainder of the CNS certainly suggests a role. T cells that specifically recognize Yo antigens have been found in the blood of PCD patients¹⁴and appear to be cytotoxic for the tumor cells. Whether this cytotoxic mechanism causes Purkinje cell loss remains to be proven.

Several other onconeural antigens have been identified by examination of serum samples from patients with paraneoplastic syndromes. Antiamphiphysin antibodies are present in serum samples of patients with stiff-man syndrome and breast cancer or, less frequently, in patients with encephalomyelitis associated with small cell lung cancer. Amphiphysin is a nerve terminal protein with a putative role in endocytosis. The antiamphiphysin antibodies of these patients predominantly react with the C-terminus of the protein that contains an SH3 domain that interacts with other proteins involved in synaptic vesicle endocytosis. The exact role of the antibodies in the pathogenesis of the stiff-man syndrome or encephalomyelitis is unknown.

A new family of paraneoplastic antigens (the Ma proteins) has recently been identified. There are at least 5 Ma proteins, the best characterized being Ma1 and Ma2. The expression of these proteins is highly restricted to neurons and spermatogenic cells of testis. In immunoblots of neuronal proteins, the anti-Ma antibodies recognize 2 bands of 37 and 40 kd that correspond to Ma1 and Ma2. These antibodies are contained in the serum and spinal fluid of patients with paraneoplastic brainstem and cerebellar dysfunction and are associated with several types of tumors (lung, breast, parotid gland, colon).¹⁵The anti-Ta antibodies are present in the serum and spinal fluid of patients with paraneoplastic limbic and brainstem encephalitis associated with testicular cancer. These antibodies recognize epitopes mainly contained in Ma2 (40-kd neuronal protein).¹⁵Some of these patients also develop mild cerebellar dysfunction. The Ma proteins are expressed by the tumors of patients with paraneoplastic syndromes but not by similar tumors from patients without paraneoplastic syndromes.

Patients with the Lambert-Eaton myasthenic syndrome develop antibodies that react with the active zones of the presynaptic cholinergic synapses, blocking the entry of calcium necessary for the release of acetylcholine. The epitopes recognized by their antibodies are contained in the P/Q-type voltage-gated calcium channels (VGCC) of the presynaptic cholinergic synapse. Similar channels are expressed in cerebellum and extracts of cerebellum labeled with iodine 125-ω-conotoxin MVIIC (toxin that specifically binds to the P/Q-type VGCC) are used in an immunoprecipitation assay to detect antibodies in patients with Lambert-Eaton myasthenic syndrome.¹⁶Antibodies against the β-subunit of the neuronal calcium channel and against synaptotagmin have also been identified in the serum of patients with Lambert-Eaton myasthenic syndrome.

Overall, these studies suggest that patients with Lambert-Eaton myasthenic syndrome develop antibodies directed against multiple epitopes, probably contained in more than 1 subunit of the P/Q-type VGCC. It is unknown whether all these antibodies are pathogenic or most of them represent elements of a repertoire of immune responses secondary to epitope spreading, with only one or a few pathogenic epitopes.

A recent study showed that 36% of patients with small cell lung cancer and PCD harbored P/Q-type VGCC antibodies in their serum; half of these patients developed clinical signs of Lambert-Eaton myasthenic syndrome.¹⁷Whether these antibodies are involved in the cerebellar disorder is unknown, but the abundance of P/Q-type VGCC in cerebellum suggests a possible pathogenic role.

Although these disorders are rare, the investigation of paraneoplastic syndromes has been fruitful for the clinical neurologist, the clinical oncologist, and the neuroscientist. The neurologist presented with a disorder of unknown cause who identifies in the patient's serum a paraneoplastic antibody can with confidence assume that the patient has cancer and notify their oncological colleague as to the likely location of a small and possibly curable neoplasm. The oncologist who encounters a patient with an antibody-positive paraneoplastic syndrome not only knows the area of the body in which to look for the tumor but also can estimate that the growth of the tumor is likely to be more indolent than the same tumor in a patient without paraneoplastic syndrome. This knowledge may have an influence on the therapy that the oncologist prescribes. The neuroscientist has at his or her disposal human serum samples containing antibodies at high titer that identify protein antigens largely restricted to the nervous system and having functions that appear to be essential for the development and maintenance of neuronal function. These antibodies can be used as probes to immunohistochemically localize the antigen in the nervous system, as well as probes to clone from complementary DNA expression libraries genes that code for onconeural antigens. In each instance in which this has been done, the antigens have become of extraordinary interest in terms of nervous system function.

Accepted for publication December 22, 1998.

Supported in part by National Institutes of Health, Bethesda, Md, grant NS26064 (Drs Dalmau and Posner). Dr Posner is the Evelyn Frew American Cancer Society Clinical Research Professor.

Corresponding author: Jerome B. Posner, MD, 1275 York Ave, New York, NY 10021 (e-mail: posnerj@mskcc.org).