Ramachandran et al. reported a surprising type of membrane-associated proteasomes designated as the neuronal membrane proteasomes (NMPs) [150,151]. As the name suggests, these proteasomes are found at pre- and post-synaptic plasma membranes in neurons, which were confirmed by immunogold electron microscopy (IEM), surface biotinylation, immunofluorescence imaging with antibody feeding and proteinase protection assays. NMPs are thought to be comprised of the 20S CP only, since no 19S components (such as Rpt5 or Rpn1) were found by IEM in these particular membrane proteasomes. NMPs are capable of degrading newly synthesized polypeptides, which are still unfolded, to short peptides. More fascinatingly, the authors showed that these peptide products could exit the cells through NMPs and be released into the synaptic cleft to function as neurotransmitters. Therefore, NMPs function not only as a protein degrader but also a new form of membrane channel to mediate cell–cell communications [152]. Although these findings were very unique and intriguing, the molecular and biochemical details of the NMPs remain unclear. First, it is curious that the 20S CP, which is soluble and hydrophilic, could be fully embedded within the hydrophobic membrane. How is the CP targeted to the plasma membrane and how does it overcome the energy barrier to traverse the lipid bilayer? It was proposed that glycoproteins, such as GPM6, could facilitate this process [151], but a clear mechanistic explanation is still needed. Second, does the NMP exhibit any substrate selectivity? The proposed role of NMPs in cleaving nascent proteins suggests that substrate availability depends on localized protein synthesis by ribosomes in the vicinity [150]. However, if the NMP complex also contained auxiliary factors yet to be identified, it might recognize and process folded protein substrates as well. On the other hand, the recent discovery that the 20S CP can by itself degrade ubiquitinated proteins [153] also implies that NMPs may have a broader range of substrates. A following question is the molecular composition and regulatory mechanisms of the NMPs. Finally, what is the function of NMP in vivo? Additionally, how can we specifically maneuver it for research and therapeutic purposes without affecting the bulk of proteasomes inside the cell? Answering these questions will depend on new technical advances in imaging, chemical biology, proteomics, structural biology and genetic models, which makes it challenging but also rewarding at the same time.

A third means of targeting the proteasome to the membrane is through lipid modification. N-myristoylation of the Rpt2 subunit has been observed by mass spectrometry in multiple species, ranging from yeast to plants to mammals [154,155,156,157,158,159,160,161]. Typically, N-myristoylation occurs co-translationally on nascent polypeptides still bound to the ribosome, where the 14-carbon saturated fatty acyl group is covalently linked to the second amino acid (almost always a Gly) after the initiator methionine is removed by methionyl aminopeptidase [162,163,164]. Notably, among all proteasome subunits of mammalian cells, Rpt2 is the only one that begins with Met-Gly, serving as the only site of the entire proteasome complex for N-myristoylation. This MG sequence of Rpt2 is strictly conserved from yeast to human, suggesting that Rpt2 is likely to be myristoylated in all species. In yeast, myristoylated Rpt2 has been shown to target proteasomes to the nuclear envelope, which is required for nuclear protein quality control [156,157]. Blocking this modification with the Rpt2-ΔG or Rpt2-G2A mutations causes mislocalization of nuclear proteasomes to the cytosol.

The role of Rpt2 myristoylation in higher organisms has not been rigorously investigated, despite Rpt2 being one of the most abundantly myristoylated proteins in human cells [161]. Our recent work demonstrated that wild-type human Rpt2 proficient for myristoylation was found at the plasma membrane, with some distribution at membrane-bound organelles as well. Membrane localization was abolished by the same ΔG/G2A mutations of human Rpt2. However, in stark contrast with results from yeast, loss of Rpt2 myristoylation in mammalian cells led to Rpt2 enrichment in the nucleus [84]. A serendipitous finding was that myristoylation-mediated membrane association is a prerequisite for Rpt2 phosphorylation at Tyr439 (Y439) by the tyrosine kinase Src, which itself is a well-established myristoylated protein tethered to the membrane [84,165]. Moreover, Rpt2-Y439 phosphorylation could be reversed by the phosphotyrosine phosphatase PTPN2 (also known as T cell PTP or TC-PTP). PTPN2 has multiple splicing isoforms. Rpt2-pY439 could only be dephosphorylated by the membrane-bound isoform of PTPN2 known as TC48, but not by the nuclear isoform TC45 [84]. Hence, the kinase, phosphatase and substrate are all placed in the same neighborhood confined by the membrane.

The biochemical consequence of Rpt2-Y439 phosphorylation is readily conceivable, as it is the very tyrosine residue within the highly conserved HbYX tail (hydrophobic residue—Tyr—any amino acid) of Rpt2 required for RP–CP association. Rpt2-Y439 is the most frequently detected pTyr site of all 19S subunits. The phosphorylation was seen in the developing rat brain but more evidently detected in cancer cells with hyperactive Src [84]. Src-mediated Rpt2-Y439 phosphorylation selectively inhibited the activity of membrane-associated proteasomes as demonstrated by a membrane-targeted reporter protein, Myr^Rpt2-GFPodc. On the contrary, the Src-specific inhibitor saracatinib/AZD0530 blocked Y439 phosphorylation and enhanced proteasomal degradation of membrane-bound substrates. Importantly, this seemed to be an integral part of the anti-cancer effects of saracatinib, since cancer cells expressing the nonphosphorylatable Y439F mutant were more resistant to this drug, both in vitro and in vivo [84]. Thus, reversible phosphorylation of Rpt2-Y439 provides a unique example of localized regulation of membrane-associated proteasomes.