Post-mortem brain autopsies were conducted on three individuals were enrolled in phase I (case 1 and 2) or phase II (case 3) trials to test the safety and efficacy of Gosuranemab for individuals with possible or probable PSP (Table 1). All three individuals were randomized to treatment arms and continued to receive Gosuranemab during open label phase extensions ranging from 10 to 30 infusions up to 1 to 2 months prior to death. No serious adverse events or obvious clinical alterations were observed. Another post-mortem brain autopsy (case 4) was conducted on an individual clinically diagnosed with PSP who was the sibling of case 2 but who did not receive Gosuranemab. While PSP is typically sporadic, this kindred provided a unique opportunity to compare the neuropathology of two affected siblings who were differentially exposed to Gosuranemab. Neither sibling had mutations in microtubule-associated protein tau (MAPT) or valosin-containing protein (VCP) which are the only known genes linked to autosomal dominant forms of primary tauopathy .
PHF1 immunohistochemistry performed on brain sections from the unimmunized sibling (case 4) detected phosphorylated tau-positive lesions with a cellular- and region-specific pattern indicative of stage 4 PSP . Specifically, the cerebral cortex including the middle frontal cortex, angular gyrus, superior temporal gyrus, anterior cingulate, and visual cortex exhibited scant numbers of tufted astrocytes, characterized by tau-positive densely packed fibrils in the proximal processes (Fig. 1a). The medial temporal lobe contained minimal neuronal tau (Braak stage 0) with a moderate density of tufted astrocytes in the amygdala. Subcortical nuclei had a relatively high burden of tau inclusions including numerous tufted astrocytes. Similarly, the brainstem showed overall severe tau pathology with globose tangles (Fig. 1b), tufted astrocytes, and oligodendroglial coiled bodies (Fig. 1c). PSP-tau aggregates contained 4R-tau isoforms (Fig. 1d), but not 3R-tau isoforms (Fig. 1e). Immunohistochemistry using GT-38, a monoclonal antibody that recognizes a conformation present in 3R + 4R AD tau, was negative in PSP-tau inclusions (Fig. 1f). These findings were indicative of a neuropathologic diagnosis of PSP.
Tauopathy associated with anti-tau passive immunization
PHF1 immunohistochemistry on immunized PSP cases (case 1 and 2) demonstrated tufted astrocytes, globose tangles, and oligodendroglial coiled bodies in a distribution and density that is typical of stage 5 (case 1) or 4 (case 2) PSP . Passive immunotherapy in these two individuals did not result in clearance of PSP-tau inclusions, corroborated by digital image analysis of the percent area occupied by PHF1 immunoreactivity from midbrain sections comparing these two immunized PSP cases (1.03% ± 0.12) with 9 unimmunized PSP cases which included the unimmunized sibling (1.06% ± 0.17; Supplemental Fig. 1). Additional semi-quantitative grades for neuronal, astroglial and oligodendroglial tau inclusions across all brain sections did not reveal appreciable differences between immunized and unimmunized PSP cases (Supplemental Table 2). For example, there were few tufted astrocytes and neurofibrillary tangles throughout all cerebral cortex regions, whereas moderate numbers of tufted astrocytes and neuronal inclusions were present in the amygdala. Tufted astrocytes were numerous in the basal ganglia (Fig. 2a) where neuronal and oligodendroglial inclusions were also observed. Brainstem structures were heavily affected by tau pathology including globose tangles (Fig. 2b), coiled bodies (Fig. 2c) and neuropil threads. GT38 staining of the medial temporal lobe demonstrated a mild burden of AD neurofibrillary tangles pathology in both immunized PSP cases (Braak stage I).
Despite the absence of phospho-tau clearance (Supplemental Fig. 1 and Supplemental Table 2), both cases demonstrated what may represent treatment-related changes. Distinct, punctate tau immunoreactivity in astrocytic processes was found frequently throughout cerebral cortex gray matter, most prevalent in the angular gyrus (Fig. 2d) and middle frontal cortex (Fig. 2e). Higher magnification revealed vesicle-like tau immunoreactivity in the astrocytic processes mostly around blood vessels (Fig. 2f and 2g). Based on their morphology and microregional distribution, we termed this type of tau-positive astrocyte “perivascular vesicular astrocytes” (PVAs). PVAs contained 4R-tau isoforms (Fig. 2h), with very few vesicular astrocytes (VAs) being detected with anti-3R-tau antibodies seen only in the middle frontal neocortex of case 2 (Fig. 2i). The GT38 antibody did not detect either the PSP-tau inclusions or PVAs (Fig. 2j).
On the basis of the morphologic and immunohistochemical profiles of astrocytic tau, we quantified neocortical astrocytic tau pathologies into the following categories: (1) perivascular vesicular astrocytes defined as vesicle-like tau immunoreactivity in the astrocytic processes around blood vessels; (2) non-perivascular vesicular astrocytes defined as vesicle-like tau immunoreactivity in the astrocytic processes not associated with a blood vessel; (3) perivascular tufted astrocytes defined as densely packed tau-immunoreactive fibers in the proximal part of astrocytic processes around blood vessels; (4) non-perivascular tufted astrocytes defined as densely packed tau-immunoreactive fibers in the proximal part of astrocytic processes not associated with a blood vessel. Blinded quantification of the middle frontal cortex demonstrated that the number of PVAs was significantly higher in the immunized group (n = 2, median = 27) compared to the unimmunized group (n = 11, median = 1, p = 0.013, Fig. 3a). There was no difference in non-perivascular VAs between the two groups (immunized group n = 2, median = 1.5; unimmunized group n = 11, median = 1; p = 0.218, Fig. 3a). Quantification of the same sections revealed no differences between immunized or unimmunized groups in the number of perivascular tufted astrocytes (immunized group n = 2, median = 6.5; unimmunized group n = 11, median = 3; p = 0.423, Fig. 3b) and non-perivascular tufted astrocytes (immunized group n = 2, median = 25; unimmunized group n = 11, median = 31; p = 0.923, Fig. 3b).
Analysis of the angular cortex yielded similar results. The quantity of angular cortex PVAs was higher in the immunized group (n = 2, median = 10) compared to the unimmunized group (n = 11, median = 0; p = 0.013, Fig. 3c). In contrast, there were no significant differences between immunized and unimmunized PSP cases in terms of the number of non-perivascular VAs (immunized group n = 2, median = 2; unimmunized group n = 11, median = 1; p = 0.346, Fig. 3c), perivascular tufted astrocytes (immunized group n = 2, median = 6; unimmunized group n = 11, median = 3; p = 0.205, Fig. 3d), and non-perivascular tufted astrocytes (immunized group n = 2, median = 30; unimmunized group n = 11, median = 19; p = 0.539, Fig. 3d). These results suggest that neocortical PVAs were associated with Gosuranemab immunization with the notable caveat that statistical interpretation is limited by the low number of available cases (Supplemental Fig. 2).
Lysosomal tau accumulation in PVAs
To identify the specific nature of the tau-positive vesicular structures in PVAs, double immunofluorescence staining was performed on middle frontal sections of immunized and unimmunized PSP cases with antibodies against PHF1 together with early endosome antigen 1(EEA1) as an endocytic marker, microtubule-associated proteins 1A/1B light chain 3B (LC3B) as an autophagosome marker, or lysosomal-associated membrane protein 1 (Lamp1) as a lysosome marker. As shown by confocal microscopy, tufted astrocytes in unimmunized PSP tau did not colocalize with EEA1, LC3B, or Lamp1-positive vesicles (Fig. 4a). Similarly, tau immunoreactivity in PVAs did not colocalize with EEA or LC3B in immunized PSP cases (Fig. 4b). In contrast, tau-positive vesicles in PVAs almost completely colocalized with Lamp1-positive vesicles (arrowheads, Fig. 4b). These findings suggest that immunotherapy-associated PVAs accumulate tau within lysosomes.
The perivascular localization of PVAs also raised the possibility that these astrocytes perhaps may be exposed to the peripherally administered Gosuranemab. To explore this possibility, double immunofluorescence microscopy showed that some PHF-1 positive vesicles in PVAs show immunoreactivity for human IgG4 (arrows, Fig. 4b) which was not seen in tufted astrocytes (Fig. 4a). While definitive localization of Gosuranemab within PVA lysosomes cannot be demonstrated here, these results raise the possibility that peripherally administered Gosuranemab may be taken up by perivascular astrocytes due to their proximity to the blood–brain barrier.
Granular/fuzzy astrocytes (GFAs) are a subtype of aging-related tau astrogliopathy (ARTAG), seen in the aging brain . GFAs are characterized as astrocytic lesions with fuzzy fibrillar or fine granular tau immunoreactivity along astrocytic processes , preferentially found in gray matter relative to white matter . These morphological and distributional features of GFAs are similar to those of PVAs in immunized PSP cases. Moreover, GFAs has been reported in primary tauopathies . Therefore, we sought to differentiate PVAs from GFAs to demonstrate the specificity of PVAs to passive immunotherapy. We identified ten non-PSP cases with neocortical ARTAG for morphologic and immunohistochemical analysis of GFAs. In the PHF1-stained neocortical sections of the ARTAG control cases, two patterns of tau-positive GFAs were detected, the fuzzy fiber-like appearance in a combination with tiny granules (Fig. 5a) and, to a lesser extent, a more granule-like appearance (Fig. 5b), both of which were predominantly found in superficial neocortical layers. However, PVAs found in the immunized PSP were predominantly granular or vesicular without tau-positive fibers (Fig. 5c) across all neocortical laminae. Furthermore, PHF1 immunoreactivity in neocortical GFAs in normal control cases did not colocalize with EEA, LC3B, or Lamp1 (Fig. 5d–f), in contrast with neocortical PVAs in immunized PSP cases which extensively colocalized with Lamp1 (Fig. 5f). Thus, morphologic and immunophenotypic features are able to distinguish between PVAs and GFAs.
Anti-tau immunotherapy and reactive gliosis
While the mechanisms by which Gosuranemab inhibits tau pathology in preclinical models is not entirely clear, the above results suggest that passive immunotherapy is associated with glial alterations. To further evaluate the glial responses associated with anti-tau immunotherapy in PSP, brain sections were immunostained for Iba1 to label microglia. Unimmunized PSP cases revealed ramified, homeostatic appearing microglia in neocortical regions with only a few scattered reactive microglia with cell body hypertrophy and thickened cell processes, consistent with the relative sparsity of tauopathy in these regions (Fig. 6, top row). In contrast, neocortical sections from both immunized PSP cases exhibited a proliferation of reactive bipolar or rod-shaped microglia with hypertrophied cell bodies and coarse processes that extended apically perpendicular to the pial surface and basally towards the subcortical white matter (Fig. 6, top row, arrowheads). Given the rarity of bipolar microglia in PSP , these observations suggest that anti-tau passive immunotherapy is associated with an atypical microglial response.
Neocortical sections were also stained with anti-glial fibrillary acidic protein (GFAP) antibody as an astrocytic marker. GFAP immunohistochemistry of unimmunized PSP cases showed no or rare reactive astrocytes across all neocortical regions (Fig. 6, second row). In contrast, there was mild (case 2) or severe (case 1) neocortical reactive astrogliosis (Fig. 6, second row) in immunized PSP cases. Notably, the reactive astrocytes observed in both immunized PSP cases exhibited an unusual bushy appearance which is different from typical reactive astrocytes which exhibit a more spidery to gemistocytic morphology.
Iba1 and GFAP immunostains were also performed on brainstem (midbrain) sections of both unimmunized and immunized PSP cases which revealed reactive microglia (Fig. 6, third row) and reactive astrocytes with hypertrophic cellular processes (Fig. 6, bottom row) in all PSP cases consistent with the high PSP-tau burden in these regions. Moreover, reactive glia appeared somewhat increased in the two immunized cases relative to unimmunized cases. Indeed, digital image analysis of the percent area occupied by Iba1 immunoreactivity showed an increase in microgliosis in immunized cases (Supplemental Fig. 3a). Evaluation of the percent area occupied by GFAP immunoreactivity in midbrain sections was difficult for several cases due to the high density of background glial processes. Therefore, we counted astrocyte numbers over multiple high-power fields based on either GFAP or Sox9 immunohistochemistry. Both immunostains revealed an increase in the number of midbrain astrocytes in immunized cases relative to unimmunized cases (Supplemental Fig. 3b–d) and were highly correlated with each other (Supplemental Fig. 3e; linear regression model, n = 13, r2 = 0.637, p = 0.001). However, counts based on GFAP immunohistochemistry were not statistically significant while counts based on Sox9 stain were statistically significant when comparing immunized and unimmunized cases, perhaps because the GFAP stain was less sensitive than the Sox9 stain (Supplemental Fig. 3f).
Anti-tau immunotherapy in corticobasal degeneration
A post-mortem brain autopsy was performed on one additional individual with a clinical diagnosis of PSP who received Gosuranemab (Table 1, case 4). Rather than identifying PSP, this case exhibited abundant tau-positive neurites, astrocytic plaques, and coiled bodies in a distribution indicative of CBD. Again, there was no clear evidence that the immunotherapy was associated with clearance of CBD-tau lesions.
Astrocytic plaques in unimmunized CBD cases consisted of tau-immunoreactive thick fibers extending into distal processes of astrocytes (Fig. 7a). Based on the above described finding of PVAs in immunized PSP cases, we determined whether similar PVAs could be seen in the case of CBD status-post immunotherapy. Although, the density CBD-tau in cerebral cortex regions consisting of threads and astrocytic plaques made identification of PVAs difficult, PVAs were ascertainable in cerebral cortex sections (Fig. 7b).
To corroborate this morphology-based assessment, double immunofluorescence microscopy of neocortex from a series of unimmunized CBD cases and the immunized CBD case was performed to differentiate astrocytic plaques and PVAs. Tau immunoreactivity in astrocytic plaques did not colocalize within EEA1 (endosomes), LC3B (autophagosomes), or Lamp1 (lysosomes, Fig. 7c). However, in the immunized CBD case, tau-positive PVAs were found which exhibited some colocalization between tau and Lamp1 lysosomal vesicles (Fig. 7d). Tau immunoreactivity in PVAs in this immunized CBD case did not colocalize with EEA1 or LC3B vesicles (Fig. 7d). These findings confirm that PVAs appear to be associated with anti-tau immunotherapy in the setting of FTLD-tau.