Table 1 Box: Validation of GVB identity

Both the core and membrane of GVBs carry epitopes that can be used to confirm a GVB identity. Commonly used GVB core markers that consistently detect GVBs in human brain and experimental models are CK1δ (Fig. 1a-d) and CK1ε [33, 67, 143], CHMP2B [90, 143, 147], pPERK (Fig. 1c-f), peIF2α and pIRE1α [43, 67, 143] and pTDP-43 [47, 59, 79, 148]. In Supplementary Table 1, an overview of primary antibodies used to detect these common GVB markers in tissue and culture is provided. Of these common GVB markers, the evidence for the localization of CK1 isoforms to human and experimental GVBs meets high standards of rigor. The presence of CK1δ and CK1ε in human and experimental GVBs has been shown using antibodies that stain GVBs in a manner independent of phosphorylation – as phosphatase pre-treatment of human brain tissue did not affect GVB immunolabeling [33]. When comparing CK1δ to CK1ε staining in aged tau Tg mice, a high degree of overlap was found: in ~ 40% of GVB-bearing neurons, all GVBs were double immunopositive, with the great majority of remaining neurons showing between 50 and 99% of double-labeled GVBs [67]. In addition, CK1δ and CK1ε staining overlap with GVBs detected by H&E staining in serial brain sections [33]. Furthermore, using STED super-resolution microscopy and/or immuno-EM the subcellular localization of CK1δ [33, 143] and CK1ε [74] in the GVB core has been confirmed at high resolution. Moreover, fluorescently-tagged CK1δ localizes to GVBs in vitro [143]. Therefore, CK1δ is currently the only constituent of the GVB core of which the presence has been confirmed directly, without the use of antibodies. In 100% [90]/98% [30] of CK1δ-positive neurons in the human brain, GVBs were also labeled by CHMP2B. On the subcellular level, CK1δ and CHMP2B co-localize in the core of 82% of GVBs in the human brain [30]. The overlap between the two GVB markers was somewhat lower in aged tau Tg mice, with CHMP2B immunolabeling being detected in 57% of neurons with CK1δ-positive GVBs and within those neurons 63% of GVBs being double positive for both CHMP2B and CK1δ [90]. In the human brain, CHMP2B staining overlaps with H&E staining after ethanol-mediated de-staining [146, 147] and mirrors H&E staining of GVBs in adjacent sections [147]. Furthermore, GVB counts based on CHMP2B and H&E staining strongly correlate [147]. Importantly, in cultured neurons CHMP2B does not only stain GVBs, but also yields a punctate staining pattern in control cells [143]. Therefore, in vitro GVB detection using CHMP2B requires co-staining with an additional GVB marker. Also the UPR activation markers pPERK, peIF2α and pIRE1α localize to structures that are morphologically clearly recognizable as GVBs in the human and tau Tg mouse brain [43, 67, 82, 99] (Fig. 1). The localization of pPERK to human GVBs was also shown by immuno-EM [82]. pPERK immunolabeling highly co-localizes with CK1δ-positive GVBs in aged tau Tg mice: in ~ 55% of neurons, all GVBs were double-positive for both markers – which is a higher percentage than found for the co-localization between CK1δ and CK1ε (see above) in single GVBs – and in the other ~ 45% of neurons between 50 and 99% of GVBs were positive for both CK1δ and pPERK. Also in the human brain and cultured neurons, stainings for the UPR activation markers overlap with CK1δ in GVBs [143] (Fig. 1), although the exact percentages of co-localization remain to be quantified. In conclusion, pPERK, peIF2α and pIRE1α are adequate markers of human and mouse GVBs. The Thal GVD neuropathological staging system [132] is based on immunopositivity for CK1δ and CK1ε and in addition pTDP-43 [59]. pTDP-43 localizes to structures that morphologically resemble GVBs in the human [59, 79, 132] and mouse brain [148], but has so far not been tested in the in vitro GVB model. pTDP-43 staining in tissue overlaps with CK1δ- and CK1ε-immunolabeled GVBs [132, 148], although the co-localization was not quantified. However, in the human brain the distribution pattern of GVBs is similar when determining the Thal GVD stage using CK1δ, CK1ε or pTDP-43 [132]. Furthermore, no difference was found when quantifying the percentage of GVB-positive neurons in the human brain using antibodies against CK1δ or the newly discovered GVB-localizing protein pMLKL of which the staining pattern overlaps with CK1δ- and pTDP-43-positive GVBs [68]. This indicates that also pMLKL is a suitable GVB marker in the human brain, whereas its use in experimental models remains to be validated. Although pPERK, peIF2α, pIRE1α and pTDP-43 (and pMLKL) are detected in GVBs using phospho-specific rather than generic antibodies which could interfere with their functional interpretation (Table 3), they are reliable and consistent GVB markers across tissue from different species and in experimental models. The GVB membrane that surrounds the core and vacuole is positive for LAMP1 in tissue and cells [30, 143] and can alternatively be detected using LIMP2 [143]. Although the global marker profile applies to most GVBs, future studies should characterize in more detail whether differential marker positivity of individual cores identifies different GVB structures and functions. Based on the existing literature on GVBs in post-mortem tissue and experimental models, we propose refinement of the criteria for validation of GVB identity in research settings, in order to systematize future post-mortem and experimental studies.

In the context of diagnostics, GVBs are typically detected by H&E staining. For research purposes, we here propose refinement of the criteria to identify GVBs to aid future studies (Fig. 1). These guidelines are of use for the validation of GVB identity in tissue and cell models. We strongly recommend that the presence of a CK1δ-positive core is a prerequisite to classify an organelle as GVB (Fig. 1a, b) as CK1δ is one of the most consistent and rigorously validated GVB core markers in tissue and cells and its GVB localization is confirmed in an antibody-independent manner. In addition, immunopositivity of the CK1δ-labeled GVB core for at least one of the other common GVB core markers (pPERK, peIF2α, pIRE1α, CK1ε, pTDP-43 and CHMP2B) is recommended, demonstrated preferably by double immunolabeling (Fig. 1c, d) or by staining of adjacent tissue sections. In addition, confirmation of the distinctive GVB morphology, including the presence of the GVB membrane that surrounds core and vacuole, is recommended. In cells as well as tissue, this can be done by visualization of the GVB membrane by immunodetection of the lysosomal membrane protein LAMP1 (Fig. 1f) or LIMP2. Alternatively, this staining could be replaced by morphological confirmation of membrane, core and vacuole on H&E staining. In addition, the GVB membrane often becomes visible upon the chromogenic peroxidase-catalyzed detection of immunolabeling with a GVB core marker – e.g. by means of the widely used chromogen 3,3′-diaminobenzidine (DAB) –, whereas upon fluorescent detection only the GVB core is shown (compare pPERK immunoreactivity in Fig. 1e with Fig. 1c, d, f). This may be due to deposition of the diffusible chromogenic product at membranes during the enzymatic signal amplification step. Therefore, as DAB signal may incorrectly suggest localization of a protein to both the GVB core and membrane, fluorescent double labeling is advised to determine the subcellular localization of GVB proteins. Lastly, super-resolution microscopy and EM can be employed to study the morphology of immunolabeled GVBs at high resolution. In conclusion, we recommend three criteria for the assessment of GVB identity that includes immunoreactivity for CK1δ (criterion 1) and an additional common GVB marker (criterion 2) in the GVB core and visualization of GVB membrane and/or vacuole (criterion 3) (Fig. 1).