Biochemically, ZS patients are characterized by the absence of functional peroxisomes and an almost complete disruption of peroxisomal beta-oxidation, leading to the accumulation of branched and very long-chain fatty acids, abnormal bile acids, and leukotrienes. In addition, the biosynthesis of docosahexaenoic acid (DHA) and plasmalogens is impaired, resulting in a drastic decrease of these important lipid components of the central nervous system. The metabolism of pipecolic acid and the peroxisomal steps of isoprenoid biosynthesis are also impaired to some extent.

Several mechanisms have been suggested:

Accumulation of branched and very long-chain fatty acids could lead to a structural disruption of cell membranes, particularly in the central nervous system (reviewed in [4]).

The disturbance of lipid metabolism could also lead to a more subtle disturbance of membrane micro-domains (rafts, caveolae) involved in intercellular signaling.

Deficiency of major myelin constituents, such as plasmalogens, DHA, and cholesterol, might lead to impaired neuronal development (see for example [5] for a review of the beneficial effects of DHA supplementation in ZS patients).

Increased levels of leukotrienes due to beta-oxidation deficiency could precipitate an inflammatory process that might be involved in the pathogenesis of the brain defects [6,7].

Studies in a mouse model of ZS have shown that reduced levels of the ether phospholipid platelet activating factor (PAF) and consequent impairment of glutamatergic signaling might cause the neuronal migration defects by an unknown mechanism [8].

The importance of lipid metabolism for the pathogenesis of ZS is underlined by the phenotype of peroxisomal single enzyme-deficient patients. Disruption of peroxisomal beta-oxidation at the level of multifunctional protein type 2 leads to the same pathological syndrome as ZS, although the patients have a slightly better prognosis [9]. Patients with acyl-CoA:dihydroxyacetone phosphate acyltransferase deficiency have rhizomelic chondrodysplasia punctata, which shares many of the key symptoms with ZS, including hypotonia, feeding problems, growth retardation, and early neonatal death [9]. On the other hand, defects of many other peroxisomal enzymes (catalase, glutaryl-CoA oxidase, phytanoyl-CoA hydroxylase, alanine:glyoxylate aminotransferase, 2-methylacyl-CoA racemase, pipecolic acid oxidase, and mevalonate kinase) as well as deficiencies in non-peroxisomal metabolism do not lead to ZS.

The most important recent data on the pathogenesis of ZS come from the analysis of three mouse models, that are deficient in different components of the peroxisomal biogenesis machinery (Pex2, Pex5, and Pex11beta, respectively [10-12]). All three mice show an almost exact phenocopy of the human ZS, including severe hypotonia at birth, impaired neuronal migration in the same brain areas, and early neonatal death. However, some important pathologies such as hepatomegaly, renal cysts, and bone abnormalities are not recapitulated. Mice disrupted in Pex11alpha are externally indistinguishable from wild-type animals [13]. Detailed analysis of these ZS models has demonstrated that DHA and isoprenoid/cholesterol biosynthesis are unlikely to be central to the pathogenic mechanism [14-16]. A more detailed study of Pex5 knock-out mice further established that changes in very long chain fatty acid, docosahexaenoic acid, or plasmalogen levels in brain are unlikely to be responsible for the neuronal migration defects [17]. Most strikingly, however, the Pex11beta knockout mouse (which has intact peroxisomes) dies of ZS, but shows almost normal peroxisomal metabolism, including normal levels of very long-chain fatty acids and plasmalogens, thus confuting the major pathogenic hypotheses listed above [12]. This conclusively establishes that the obvious biochemical disturbances are concomitant but not causal to the most serious symptoms of ZS patients.

In addition, analysis of a mouse model of peroxisomal beta-oxidation deficiency indicates that the most severe ZS-like symptoms are independent of the characteristic neuronal migration defects [18,19]. Figure 1 summarizes the molecular conditions leading to Zellweger-like syndromes in human and mouse.

Figure 1 Molecular conditions leading to Zellweger-like syndromes in human and mouse. Biochemical and pathological consequences that are not common to all disorders are indicated. PEXn = any of the peroxisome biogenesis proteins that can be mutated in classical (more ...)

We decided to use microarray data to identify possible players involved in such a misregulation mechanism in human ZS. Microarrays provide an unbiased survey of gene expression levels in human tissues and cells under various experimental conditions. Many hundred data sets, examining most of the predicted genes of the human genome, are now publicly available for exploratory analysis. Based on the known clinical data of ZS patients, we defined three criteria for candidate genes:

1. A candidate should be regulated positively or negatively in concert with lipid metabolism genes, because changes in acyl-CoA derivatives are the only common factor in ZS and the ZS-like single enzyme deficiencies.

2. To be active in a regulatory pathway, a candidate should be involved in a signaling pathway or in transcriptional regulation, so that feedback regulation is possible.

3. Because the most serious symptoms of ZS affect the brain, an ideal candidate would be involved in brain or neuron function.

Using these criteria we identified only a single candidate gene, Diazepam Binding Inhibitor, DBI, in all data sets available in the Stanford Microarray Database http://genome-www5.stanford.edu/MicroArray/SMD/. Co-regulation of DBI with lipid metabolism is found consistently in experiments representing the two current microarray methodologies, Affymetrix oligonucleotide arrays [21] and spotted two-color cDNA arrays [22] (Table 1). DBI is also among the most highly upregulated genes in lovastatin-treated squamous cell carcinomas [23].

Table 1 Ten genes that have expression profiles most similar to diazepam binding inibitor, DBI, in two large-scale microarray expression profiling studies, as quantified by Pearson correlation coefficient r (the top 10 neighbors of DBI are shown for each sample (more ...)

In addition to GABA(A) receptor, DBI and its processing products bind to the ubiquitous mitochondrial benzodiazepine receptor pBR and play a crucial role in the acute stimulation of steroidogenesis [36-38]. A third type of DBI receptor mediates rapid release of intracellular calcium, but remains to be characterized in detail [39,40].

In its role as an acyl-CoA binding protein, DBI specifically binds to medium and long-chain acyl-CoA esters (but not to fatty acids, acyl carnitines, cholesterol and a number of nucleotides; [41]) and is involved in fatty acid-mediated regulation of gene expression and the formation of intracellular acyl-CoA pools (reviewed in [42]).

In vitro studies have identified functional sterol response elements (SREBP binding sites) and peroxisome proliferator gamma response elements (PPARgamma binding sites) in the DBI gene promoter [43,44].

Finally, DBI has a stimulating effect on the pancreatic cholecystokinin and insulin system [45,46], that might be related to the pancreatic islet hyperplasia seen in many ZS patients.

Table 2 Comparison of selected symptoms in "fetal benzodiazepine syndrome", Zellweger syndrome and peroxisomal beta-oxidation (multifunctional protein-2) deficiency.

The work by Laegreid et al. was heavily criticized for its epidemiological validity (e.g. [52]) and later studies did not find a significant correlation between benzodiazepine use during pregnancy and birth defects (reviewed in [53]). However, some reports have linked diazepam to oral clefts, distal limb defects, microcephaly, and cardiovascular anomalies (reviewed in [53]). In addition, it is well established that high doses of diazepam lead to craniofacial malformations in rodents [54-57] and mice carrying targeted disruptions of the GABAergic pathway consistently develop a cleft palate, a rather common feature of Zellweger patients [58,59].

We are not aware of any other drug-induced teratogenic syndrome that mimics ZS. It is quite unlikely that all of Laegreid's patients just had a mild form of ZS, as they did neither show the characteristic biochemical abnormalities nor the disease progression with age. Also, consistent with the fact that benzodiazepine exposure is largely terminated at birth, they did not show the neonatal seizures that affect ZS patients.

We suggest the following tentative mechanistic model for the pathogenesis of DBI misregulation in ZS (figure 2):

Figure 2 Possible regulatory network leading to disturbed GABAergic signaling in peroxisome-deficient patients. Arrows indicate the predicted direction of change in the patients. See text for details.

Peroxisome deficiency leads to an accumulation of peroxisomal metabolic intermediates (metabolite X), most likely acyl-CoA derivatives that are also disturbed by the single enzyme defects that cause ZS-like syndromes. Directly or indirectly, metabolite X causes an inactivation and down-regulation of DBI and a decrease in DBI release, possibly by replacing its physiological acyl-CoA ligands. Decreased activity of DBI affects two intertwined feedback loops. In the first loop, decrease of DBI releases the inhibition of GABA(A) receptor [60]. Overactivation of GABA(A) receptor inhibits glutamate release [61] and consequently causes a decrease in NMDA receptor-mediated glutamatergic signaling [62]. This in turn causes a lack of calcium release in the target cells and a deficiency of DBI secretion and expression [63]. The second loop starts by a decrease in DBI action on its non-GABA(A) receptor and a subsequent lack of DBI-stimulated calcium release [39,40,64]. Both pathways converge on the decrease in intracellular calcium, that results in an inhibition of DBI release [65] and closes the loop. A hypothetical ancillary loop could involve positive crosstalk between the over-activated GABA(A) receptor and its counterpart GABA(B), which is known to inhibit DBI release [66]. The two loops are not necessarily present in the same cells, and each of them may involve several adjacent cells. Also, different processing products of DBI act on intracellular calcium using different pathways [67], and may lead to additional differentiation of the DBI misregulation effect. The main feature is that both loops tend to enhance the down-regulation of DBI, thus amplifying and maintaining a small initial disturbance caused by the peroxisomal defect. Both loops involve a deficiency of NMDA receptor-mediated calcium release, in agreement with observations in ZS mouse models [8]. Gressens et al. (2000) found that administration of GABA at a concentration of 0.25 mg/kg twice a day had no effect on neuronal migration in peroxisome-deficient and control mice. This finding might indicate that the neuronal migration defect is mediated by the direct effect of DBI-derived peptides on intracellular calcium through its non-GABA(A) receptors.

Misregulation of DBI and the consequent disinhibition of GABAergic signaling should lead to diverse defects in various brain areas. A region of special interest is the inferior olive, which is malformed or absent in classical ZS. Olivary axon collaterals do not only innervate non-GABAergic neurons in the cerebellar nuclei, but also GABAergic nucleo-olivary cells, thus establishing a direct feedback loop to the inferior olive [68]. Purkinje cells, which show characteristic heterotopias in the cerebellum of ZS patients are also GABAergic [69,70].

Furthermore, it has been demonstrated recently that excessive GABA(A) receptor activation by neurosteroids or benzodiazepines initiates a slow form of neuronal death in cultured hippocampal neurons [71]. This could be the basis of the increased neuronal apoptosis observed in ZS. It should also be noted that the suggested mechanism for ZS brain disturbances bears conceptual similarity to that for fetal alcohol syndrome, where a combination of NMDA blockade and GABA(A) over-activation leads to widespread apoptosis in the developing brain [72].

The observation of cleft palate in GABA signaling-deficient mouse models indicates that disregulation of the GABA(A) pathway might also be responsible for the malformation syndromes observed in ZS patients and in benzodiazepine embryofetopathy [58,59].

DBI is an especially suitable candidate for ZS pathogenesis because it is shown to integrate signals from a variety of lipid-related pathways (fatty acids, sterols, steroids). Its restricted binding properties narrow down the search for a specific metabolite X that is responsible for the initial repression of DBI, and the feedback loops explain how even a slight disturbance that is common to all ZS-like syndromes can tip the balance towards a disinhibition of the GABAergic system, resulting in the severe disorders that are found in all patients.

The recent availability of several peroxisome-deficient mouse models provides an excellent opportunity to directly test the GABA/DBI model of Zellweger pathogenesis. This would involve a detailed analysis of DBI isoform levels in wild type and mutant brains, as well as a study of GABA agonist and antagonist effects, both in vivo and in neuronal explants. Additional in vitro studies could be used to determine the molecular details of DBI regulation and processing, to test the existence of the feedback loops postulated by the model. After the general mechanism has been established, identification of "metabolite X" would be crucial and could be achieved, e.g., by examining the effect of fractionated brain extracts on a DBI reporter cell line.

Currently, no successful therapy for peroxisome deficient patients is available. Identification of a GABA component to the pathogenesis of Zellweger syndrome has obvious implications for the treatment of affected patients. Straightforward pharmacological intervention with well-known GABA antagonists could be used to counteract the misregulation and to interrupt the pathological feedback loop. In addition, even if the molecular details of the proposed mechanism turn out to be incorrect, consideration of an amplification of minor metabolic changes by positive feedback changes could influence the search for "metabolite X", which by implication of the present model would be a low-concentration intermediate that binds DBI, and hence is probably a Co-enzyme A thioester.

We suggest that peroxisomal deficiency leads to an unidentified disturbance of lipid metabolism that causes misregulation of the diazepam binding inhibitor, DBI. This misregulation affects two independent systems and is amplified by specific feedback loops: The DBI/GABA(A) system and the DBI/calcium system. The independence of the two systems agrees with the observation in peroxisomal beta-oxidation-deficient mice, where hypotonia and early death occur without neuronal migration defects. The activation and subsequent escalation of feedback loops would explain why several defects with a widely varying biochemical background lead to the same complex of symptoms.

Our rather simplistic model necessarily ignores many details that complicate the picture, e.g. we do not consider the regional and developmental heterogeneity in the GABAergic system (e.g. excitatory GABA action in early development and the enormous plasticity of the central nervous system that will result in important compensatory regulation phenomena. Also, relatively little is known about the functional distribution of DBI in the central nervous system and its autocrine vs. paracrine effects.

In fact, the DBI hypothesis is not able to explain all observations in peroxisome deficiencies. However, it seems that the DBI hypothesis is able to explain a larger part of the observations than previous ideas, e.g. the Pex11beta k.o. phenotype seems to follow naturally from the suggestion that slight initiating changes in a metabolite might lead to an escalating response via feed-back mechanisms. It is very well possible that the DBI hypothesis is incomplete and/or over-simplified (for example, explaining some differences between Zellweger and Zellweger-like syndromes might require postulating different mechanisms of action for DBI in these conditions) but we hope that even a partially correct hypothesis may be able to point research into a useful direction, e.g. concerning the importance of small perturbations amplified by misactivated feedback mechanisms.

The existence of several mouse models of ZS should allow for rapid testing and refinement of our idea that misregulation of DBI is involved in the pathogenesis of Zellweger syndrome.

None declared.

The pre-publication history for this paper can be accessed here:

http://www.biomedcentral.com/1471-2431/4/5/prepub

I thank Dr. D. Höller and Dr. S. K. Krisans for helpful discussions. This work was supported in part by National Institutes of Health grants DK58238 and DK58040.