Subcellular organization of UBE3A in human cerebral cortex

Protein degradation through the ubiquitin-proteasome pathway (UPP) is important for a wide range of functions in the nervous system [1, 2]. In UPP processing, the protein substrate is marked by covalent attachment of the small protein ubiquitin for subsequent degradation by the proteasome, a specialized proteolytic complex. The process starts with the activation of ubiquitin by E1 ubiquitin ligase enzymes [3, 4]. E2 enzymes then transfer the activated ubiquitin to E3 ubiquitin ligases, which attach the ubiquitin to the substrate. The E3 ligases determine the substrate specificity of ubiquitination.

Two major classes of E3s have been defined: HECT (homologous to E6-AP carboxyl-terminus) domain E3s and RING finger E3s. UBE3A, the first E3 discovered in the HECT class [5], is of special interest, because failure to maintain appropriate levels of UBE3A in the brain is directly linked to severe clinical disorders. The loss of UBE3A function causes Angelman syndrome (AS) [6, 7], characterized by profound intellectual and developmental disabilities [8]. On the other hand, increased UBE3A activity (via gene duplication or gain-of-function mutation) is implicated in several forms of autism [9‐15], implying that UBE3A must be expressed at precise levels to allow proper neuronal function.

Despite its clinical importance, our current understanding of UBE3A’s function in the brain remains limited. UBE3A has been implicated in regulation of Golgi acidification [16] and mitochondrial function [17‐19], as well as synaptic transmission and plasticity [20‐26]. Several potential interactors and substrates for its ubiquitin ligase activity have been identified, including the RhoA guanine nucleotide exchange factors Pbl/ECT2 and Ephexin5 [24, 27, 28], RAD23A, (an enzyme involved in excision repair of damaged DNA [29]), and the estrogen receptor [30]. UBE3A has also been shown to regulate the levels of Cbln1 in excitatory neurons, which in turn can influence sociability [26]. Finally, UBE3A can interact directly with the proteasome to influence its degradative activity [15, 31‐35], suggesting potentially wide-ranging roles for UBE3A in cellular functions that remain to be fully defined.

Crucial for the elucidation of UBE3A’s functions is to determine where the protein actually lies within the cell. Previous immunohistochemical studies in rodents found a broad distribution of UBE3A, but emphasized its nuclear expression in mature neurons, consistent with a proposed role for UBE3A in the regulation of transcription [26, 36‐38]. However, patterns of UBE3A expression in humans may not be fully conserved, considering the 100 million years of divergent evolution between rodents and primates. Unique patterns of UBE3A expression within the cerebral cortex could have important functional implications, especially given the cognitive manifestations of both AS and autism. Here, we explore this intriguing possibility by investigating the subcellular distribution of UBE3A in the human temporal cortex.

The overall pattern of staining was consistent across biopsy samples from all 9 subjects. UBE3A was found in neurons throughout the depth of cerebral cortex (Fig. 3a–c). Staining within neurons concentrated in the nucleus, but could also be seen in somata and proximal dendrites, and punctate staining was observed in neuropil (Fig. 3d, e). The apparent lack of nuclear staining in some neurons was interpreted as an artifact arising from limited antibody penetration, since it was not seen in neurons whose nucleus lay at the section edge. Examination of thin (< 300 nm) sections showed that much of the nuclear staining was organized into discrete tiny aggregates of DAB reaction product distributed through the nucleus (Fig. 3f).

GABAergic deficits have been implicated in the pathogenesis of neocortical hyperexcitability and epilepsy in AS model mice [25, 48], and GABA transporter-1 has been proposed as a potential UBE3A substrate [49]. To determine whether UBE3A is expressed in inhibitory neurons of the human neocortex, as it is in mice, we performed double labeling with GABA (Fig. 4). We found that GABAergic neurons, like pyramidal neurons, express higher levels of UBE3A in their nuclei than in their somata. However, quantitative analysis of biopsy material from three subjects revealed that the concentration of nuclear UBE3A is ~ 40% (38 ± 6.4%, n = 51) higher in pyramidal neurons than in GABAegic neurons. In addition, colocalization was observed in fine processes and puncta distributed through the neuropil (Fig. 4c, d).

In mice, UBE3A is expressed not only in neurons but also neuroglia. To evaluate its distribution in non-neuronal cells in the human temporal cortex, we performed double staining with Olig2, to identify oligodendrocytes (Fig. 5a, b). Indeed Olig2-positive somata were consistently immunopositive for UBE3A, though to a lesser degree than seen in neurons. We also performed double staining with GFAP to identify astrocytes, finding that astrocytes also expressed UBE3A, though generally at lower levels than either neurons or oligodendrocytes (Fig. 5c, d). These patterns were reminiscent of UBE3A staining of neuroglia in the mouse brain [43].

To get a better understanding of the subcellular organization of UBE3A, we performed high-resolution immunofluorescence microscopy. In images prepared using an Airyscan confocal microscope (Fig. 6), UBE3A staining appeared as discrete puncta whose density was high in the nucleus and also present (though at lower densities) in somata and proximal dendrites; scattered puncta of immunolabel were seen in neuropil (Fig. 6a). In the nucleus, UBE3A puncta were interspersed with DAPI “hotspots,” but did not overlap with them (Fig. 6b, c). Likewise, UBE3A staining seemed to be excluded from nucleoli (Nu, Fig. 6b). To more directly test the relationship between UBE3A and DNA condensation, we computed the intensity correlation quotient [47] (ICQ, computed such that − 0.5 represents perfect negative covariance and 0.5 represents perfect positive covariance between two antigens) for 92 nuclei from 3 humans Q1010, Q1011, and Q1012 (Fig. 7). We found an average ICQ of − 0.04 ± 0.005, showing that UBE3A and DAPI staining negatively co-vary, confirming our qualitative impression that UBE3A is excluded from regions of condensed DNA.

We used electron microscopy to examine the subcellular organization of UBE3A with higher precision, using pre-embedding immunogold labeling followed by silver intensification (Fig. 8). We focused our attention on layers II–III as we had done previously for mouse [41]. Labeling was prominent in nuclei but was also present in somatic cytoplasm and neuropil, echoing the results from light microscopy (Fig. 6). In the nucleus, UBE3A labeling was largely restricted to euchromatin domains; very little labeling was found over electron-dense heterochromatin zones (Fig. 8a, b), as was predicted by our ICQ analysis of UBE3A/DAPI colocalization (Fig. 7). UBE3A was also excluded from nuclear organelles such as nucleoli. Occasional gold/silver particles were associated with the nuclear membrane. In neuronal cytoplasm, UBE3A was seen decorating the Golgi apparatus and was associated with the outer mitochondrial membrane (Fig. 8c, d). In dendrites, labeling was often associated with the plasma membrane and with poorly defined tubulo-vesicular endomembranous structures. In dendritic spines, labeling was within the spine head and neck, including the spine apparatus, but was not found at the PSD (Fig. 8e). The majority of labeling within the neuropil was found in axon terminals making axospinous, axodendritic, and axosomatic synapses (Fig. 8e, f). In axon terminals, UBE3A associated with the plasma membrane and over clusters of presynaptic vesicles, but did not show a special affinity for the active zone. Double labeling experiments showed that UBE3A is expressed in GABAergic terminals (Fig. 8g).

While previous studies have examined human UBE3A expression using Western blot analysis [50, 51] and in human-induced pluripotent stem cells [52, 53], the present work provides the first high-resolution description of the subcellular distribution of UBE3A in the human brain. Though reporting only the presence of a molecule, not its functional level of activity, immunohistochemistry can shed light on important aspects of human neurophysiology and neuropathology inaccessible by even the most sensitive contemporary methods of live brain imaging. However, immunohistochemistry may be misleading unless tissue of excellent quality is studied. The potential for error is particularly troublesome when tissue of poor quality is examined using super-resolution methods. Several indirect methods are customarily used to assess the quality of human histological material including tissue pH and specific markers of RNA quality [54]. We used electron microscopy to assess tissue quality directly, finding that histological and immunocytochemical staining at the light microscopic level can appear to be satisfactory even in material that has been severely degraded, potentially resulting in misleading results.

UBE3A may also serve important functions in glia. For example, flies with glia-specific overexpression of Dube3a (the fly UBE3A homolog) display a robust seizure phenotype, suggesting a possible role for glial UBE3A in Dup15q pathophysiology [55]. Our finding that UBE3A is also expressed in glia in the human brain (though at a lower level than in neurons) is consistent with this possibility.

We found that UBE3A is expressed in multiple subcellular compartments. In neurons, it concentrates in euchromatin-rich domains within the nucleus, where it is positioned to regulate the expression of active genes, possibly exerting cell-wide global effects. A recent study shows that increasing UBE3A in the nucleus downregulates the glutamatergic synapse organizer Cbln1 [26], leading to weakened glutamatergic transmission; the same study found impaired sociability in these mutant mice, providing a possible explanation for how UBE3A overexpression might lead to autism spectrum disorder. Nuclear UBE3A may also perform additional roles. Like other ubiquitin E3 ligases, UBE3A may modulate the abundance and activity of transcriptional activators via its ubiquitin-protein ligase activity, thus modulating the formation of transcriptional complexes at genes and regulating chromatin structure [56]. UBE3A might also act more directly within the nucleus as a transcriptional co-regulator [36, 57‐59].

UBE3A labeling was also associated with mitochondria. Loss of UBE3A has been linked to key enzymatic deficits in the mitochondrial respiratory chain [17, 18]. A role for mitochondrial abnormalities in the pathogenesis of autism spectrum disorders has also been suggested [60], and mitochondrial dysfunction has been shown in autistic patients with 15q inverted duplication [61].

The enrichment of UBE3A that we observed in both excitatory and GABAergic axon terminals and (to a lesser extent) in dendritic spines is of special interest since both Dup15q and AS mouse models show synaptic deficits [13, 25, 62]. Postsynaptically, abnormal function of excitatory synapses in UBE3A mutant mice has been linked to decreased UBE3A-mediated proteasomal degradation of the RhoA guanine exchange factor Ephexin5 and to pathways modulated by the synaptic protein Arc [23, 24, 58, 63]. UBE3A has also been shown to target the small-conductance potassium channel SK2 for degradation. Decreased NMDA receptor activation due to an elevated SK2 protein level may underlie the impaired LTP seen in AS model mice [62]. Presynaptically, the localization of UBE3A over vesicle clusters is surprising, since no integral synaptic vesicle proteins have yet been identified as targets for UBE3A [64]. However, UBE3A does appear to modulate the degradation of α-synuclein, a protein closely associated with synaptic vesicles [65‐68], and loss of UBE3A can alter rates of synaptic depression [25, 48]. Although the mechanistic basis of UBE3A-associated synaptic disruption remains unclear, our data support a role for UBE3A in regulating synapses in the human neocortex.

The overall pattern of UBE3A immunoreactivity we found in human temporal cortex is generally consistent with that seen in previous studies of rodent cortex, reinforcing the idea that mice offer a valuable model in which to study the role of UBE3A in the brain. Mouse models of AS display many Angelman-like phenotypes, including learning and memory deficits, motor dysfunction, and seizures, while transgenic mice expressing UBE3A at increased dosages manifest several core autistic behavioral traits [26]. Nevertheless, some divergence between mouse and human UBE3A expression is likely. In particular, regulatory elements may have evolved in humans to sculpt UBE3A expression (which is spatiotemporally ubiquitous in the mouse) into unique and discrete patterns across the neocortex and throughout the cerebrum during development [69]. This remains to be tested in diverse samples of the developing and adult human brain.

That UBE3A is expressed at high levels in both excitatory and inhibitory neurons in the human cerebral cortex points to the many ways that its dysregulation might disrupt circuit function, while its concentration at synapses and in nuclei offers intriguing functional clues. UBE3A in axon terminals presumably regulates the function of individual synapses; in contrast, its concentration in euchromatin-rich nuclear domains suggests that UBE3A may also mediate global effects on neuronal physiology, by regulating gene transcription. This information identifies axon terminals and nuclei as two distinct potential targets for future clinical intervention.