Introduction
The semantic variant of primary progressive aphasia (SV PPA), also known as semantic dementia [
1], is a language-based neurodegenerative disorder which presents as a fluent, anomic aphasia with single-word comprehension problems and generally progressive loss of vocabulary and non-verbal knowledge [
1,
2]. SV PPA has been associated with bilateral atrophy of the ventrolateral anterior temporal lobe, which is usually more prominent in the dominant language hemisphere [
3]. In 69–83% of cases, the underlying pathology is frontotemporal lobar degeneration (FTLD) transactive response DNA-binding protein TAR DNA-binding protein-43 (TDP-43) type C pathology [
4,
5]. Alzheimer’s disease (AD) pathology occurs in up to 20% of clinical cases and more rarely Pick’s disease (PiD) is seen as underlying pathology [
6]. Currently, a positron emission tomography (PET) tracer to visualize TDP-43 in vivo is not available yet. However, first-generation tau-PET tracers such as [
18F]AV1451 [
7] and [
18F]THK5351 [
8] have consistently shown strong in vivo tracer retention in the anterior temporal lobe of SV PPA patients [
9‐
16]. At the individual patient level, evidence for consistently elevated [
18F]AV1451 binding has been found in the anterior temporal lobe of all seven SV PPA cases and all seven “right” semantic dementia cases [
10,
11]. In a separate study, this pattern was also observed in all 13 SV PPA cases who were amyloid-negative [
14]. Elevated binding in this peak region has also been consistently observed with [
18F]THK5351 across SV patients [
13,
15,
16]. These findings are surprising since the vast majority of SV PPA patients have underlying TDP-43 proteinopathy [
5] and are not expected to show [
18F]AV1451 binding in the anterior temporal lobe as the ligand has selectivity for tau over amyloid (29 fold) [
7].
In vitro binding studies with fluorine-18 labeled and tritiated AV1451 on sections with FTLD pathology have shown heterogeneous results [
12,
17‐
21]. [
3H]AV1451 autoradiography indicated no binding to TDP-43 type C in SV PPA nor to C9orf72 gene mutation carriers with underlying TDP-43 type B pathology [
12]. Similar absence of binding was observed with [
18F]AV1451 phosphor screen autoradiography in TDP-43 type A [
22]. Absence of in vitro binding in cases with underlying TDP-43 has also been reported by an independent study [
21]. On the other hand, Sanders et al. and Lowe et al. suggested minimal binding to TDP-43 and strong binding to tau, in particular to AD tau [
20,
23].
We hypothesize that the elevated in vivo [
18F]AV1451 PET signal consistently observed in the anterior part of the inferior temporal/occipitotemporal gyrus in SV PPA is possibly related to TDP-43 binding or to MAO-B affinity. As indicated by a recent [
3H]AV1451 autoradiography study on brain homogenates devoid of tau pathology, the high in vivo [
18F]AV1451 retention in SV PPA might be related to binding to monoamine oxidase (MAO) A and B enzymes [
24]. [
18F]THK5351 competition autoradiography also showed affinity for MAO-B on AD sections [
25]. MAO-B is abundantly expressed on reactive astrocytes which accompany neurodegeneration [
26]. Expression of MAO-B by reactive astrocytes can be visualized by specific MAO-B PET tracers including [
11C]deuterium-L-deprenyl [
27].
Based on these findings, we pursued the hypothesis that [18F]AV1451 and [18F]THK5351 binding in SV PPA is related to affinity for MAO-B enzymes. We conducted an in vitro autoradiography binding study with [18F]AV1451 in cases with a clinical diagnosis of SV PPA, constituting the range of underlying pathologies (i.e., TDP-43 type C, AD, and PiD) and compared binding to controls and cases with typical AD, corticobasal syndrome (CBS) due to neuropathological AD, and FTD behavioral variant (FTD-bv) due to TDP-43 type C. Coincubation with the MAO-B inhibitor deprenyl (selegiline) was performed in all cases to study MAO-B binding. A subset of cases was assessed with [18F]THK5351 to assess translatability of findings to another first-generation tau-PET tracer.
Discussion
We and others have previously demonstrated strong PET signal of the first-generation tau-PET tracers in the anterior temporal lobe that was consistently present in patients with SV PPA [
9‐
16]. In an effort to explain this unexpected but highly consistent finding, we conducted an in vitro autoradiography binding study with [
18F]AV1451 in SV PPA cases, constituting the range of underlying pathologies. Binding was compared with negative controls and to typical AD, corticobasal syndrome (CBS) due to neuropathological AD, and FTD behavioral variant (FTD-bv) due to TDP-43 type C. In all three cases with SV PPA due to FTLD-TDP, no specific [
18F]AV1451 binding was observed. The absence of binding in controls as well as the successful blocking with authentic AV1451 in cases with tauopathy suggested the specificity of the [
18F]AV1451 signal for tau. The specific [
18F]AV1451 binding was highest in AD tau, followed by PiD tau. This binding colocalized with the respective tau lesions on immunohistochemistry. A subset of cases was assessed with [
18F]THK5351, indicating similar results.
The absence of binding in the vitro autoradiography images of FTLD-TDP cases is in line with a recent study in which in vitro [
3H]AV1451 autoradiography on frontal and temporal cortical cryosections of SV PPA cases as well as
C9orf72 TDP-43 type B cases also showed no binding [
12]. Another independent study demonstrated the absence of in vivo [
18F]AV1451 PET signal during life in a patient with the
C9orf72 mutation who was neuropathologically diagnosed as TDP-43 type B [
21]. Incidental co-pathology of scattered NFTs in the middle frontal and inferior temporal gyrus showed corresponding mild [
18F]AV1451 binding but without additional uptake matching the widespread TDP-43 type B pathology [
21]. Apart from two earlier studies suggesting low binding to TDP-43 type A and C [
20,
23], these findings, together with the current results, suggest that [
18F]AV1451 does not bind to TDP-43 aggregates in FTLD-TDP. Although, as demonstrated in Tsai et al. 2018 [
21], it cannot be excluded that some FTLD-TDP cases might have some tau pathology; the anterior temporal lobe localization of PET signal does not correspond to the expected distribution of tau in AD [
31].
Moreover, in an independent study, strong anterior temporal lobe [
18F]AV1451 binding was found in all seven cases with either SV PPA or “right” semantic dementia [
11] despite four of these being AD-biomarker negative. In another study, all seven SV PPA cases had elevated anterior temporal lobe [
18F]AV1451 binding but only one case appeared to be amyloid-positive on PET [
10]. A recent in vivo PET study demonstrated that when SV PPA cases were stratified based on amyloid status, all 13 amyloid-negative SV PPA cases showed increased left anterior temporal cortex [
18F]AV1451 binding, which was higher than the signal in typical AD [
14]. If amyloid-positive, the SV PPA cases also showed, besides peak binding in left anterior temporal cortex, more widespread cortical binding than the amyloid-negative SV PPA cases [
14]. In these studies, no post mortem information was available on these cases. Nevertheless, it seems highly unlikely that the amyloid-negative SV PPA cases are characterized by underlying AD based on the available amyloid-PET data.
Given that the degree of frontotemporal atrophy in FTD is related to the degree of astrocytic apoptosis [
26], we expected that SV PPA with FTLD-TDP would be characterized by extensive astrocytic apoptosis and astrogliosis. The latter becomes the overwhelming pathological feature as the disease progresses [
26]. More specifically, we expected that the strong [
18F]AV1451 signal seen in vivo would relate to underlying astrogliosis, overexpressing MAO-B [
42]. Semi-quantitative assessment of pathology demonstrated, however, that all patients included in this study (including all SV PPA) showed a similar degree of neuroinflammation (i.e., astrogliosis and microgliosis), regardless of the underlying neuropathological diagnosis. Moreover, no clear overlap was seen between neuroinflammatory markers and specific cortical [
18F]AV1451 signal on the adjacent cryosection. This leaves us with an open question regarding the binding target that is present in vivo but possibly no longer active in postmortem sections, e.g., mediated by reactive astrocytes or microglia. We speculate that target binding may require living astrocytes or microglia, e.g.,if transmembrane transporters or enzymatic modifications are involved. However, knowledge on the specific nature of these other binding targets of AV1451 is currently limited. Alternatively, the target may be present in the inflammatory reaction during an early disease stage but not in the end stage of the disease. For instance, in the AD dementia stage, neocortical in vivo binding of [
11C]deuterium-L-deprenyl is not increased compared with controls, while in the MCI stage [
11C]deuterium-L-deprenyl binding is significantly increased [
27]. We used patient tissue of cases that are end-stage disease. Therefore, the absence of MAO-B binding might not only depend on the spatial pattern of normal MAO expression [
43] but also on the temporal dynamics of disease-related processes such as gliosis. In a post mortem study using AD brain tissue, significantly higher binding was observed in the temporal lobes and in the white matter, which coincided with the presence of an increased number of activated astrocytes (on GFAP immunohistochemistry). Critically, the highest binding was observed in Braak NFT stage I-II, whereas it decreased with increasing Braak NFT stages [
44]. A similar mechanism might occur in SV PPA in vivo, in which there could be an increase in MAO-B expression associated with astrogliosis, followed by a decrease in MAO-B expression when end stage disease has been reached. Like [
18F]AV1451, [
18F]THK5351 PET also shows a very strong anterior temporal signal in SV PPA in vivo [
13], which co-localizes with atrophy [
10,
13], indicating it might be relevant to the pathophysiology associated with SV PPA.
Another source of potential off-target binding besides MAO, might relate to neuromelanin-containing cells, calcifications, or iron deposits as previously reported [
20,
22,
24,
45,
46]. However, we could not demonstrate an overlap between autoradiographic binding and iron deposits in the current study.
Furthermore, we should be aware that in vivo settings might differ from in vitro settings. For example, in Parkinson’s disease, use of MAO-B inhibitors at pharmaceutical levels did not significantly affect [
18F]AV1451 binding in vivo [
47]. This made the authors conclude that MAO-B does not appear to be a significant binding target of [
18F]AV1451, despite tau levels being low in this patient group in the first place. Similarly, we also could not reveal any binding in vitro while in vivo, a strong signal has been consistently shown [
9‐
12,
14]. On the other hand, for [
18F]THK5351, in vivo reduction of tracer uptake has been demonstrated in AD and PSP using an oral dose of selegiline [
25]. In in vitro autoradiography studies, deprenyl at concentrations of 150 nM and 500 nM, was able to reduce [
18F]THK5351 binding on AD sections of the striatum, but also on sections of the prefrontal cortex and hippocampus [
25]. We did not observe a significant effect of coincubation with deprenyl on [
18F]AV1451 binding in the AD cases. The anterior temporal lobe has generally low MAO-B expression levels in healthy conditions [
43]. Therefore, it cannot be excluded that other regions apart from the anterior temporal lobe would have shown an effect of deprenyl in AD. As discussed earlier, another caveat is that in our study, MAO-B levels might be generally low [
44] since the selected AD and PiD cases were end-stage disease.
Differences in results can also relate to the use of brain sections as used here versus brain homogenates as used in some of the other studies. When brain sections of cases that demonstrated specific binding (i.e., AD and PiD cases) were incubated with a mixture of high molar activity [
18F]AV1451 and a high concentration of the MAO-B inhibitor deprenyl, no reduction in tracer binding was observed in the current study. This indicates that MAO-B binding cannot explain the source of off-target binding in these cases. In contrast, a previous in vitro study of Lemoine et al. demonstrated a reduction in [
3H]deprenyl binding in homogenates of AD patients when a competition assay was performed with unlabeled AV1451 [
48]. In contrast to the current in vitro autoradiography study, binding studies with [
3H]AV1451 on AD brain homogenates showed 10% displacement by the MAO-B inhibitor deprenyl and 80% (
Ki = 0.70 nM) by the MAO-A inhibitor clorgyline [
24]. Also in non-AD brain frontal cortex homogenates, [
3H]AV1451 binding was completely inhibited with nanomolar affinity (
Ki = 0.43 nM) by the MAO-A inhibitor clorgyline [
49]. The [
3H]AV1451 binding affinity for MAO was comparable to its affinities for aggregated tau [
24] and as such could lead to a confounding signal in PET studies. Of note is that the latter study showed that there might be regional differences in these off-target proteins. In the temporal cortex, the [
3H]AV1451 binding is essentially sensitive to clorgyline but not to deprenyl, indicating a majority of binding to MAO-A in the temporal cortex. Given that the temporal cortex is high in MAO-A levels [
43], [
18F]AV1451 may also bind to MAO-A in vivo [
24,
50]. However, since no binding was observed in the SV PPA cases of our autoradiography experiment, the MAO-A binding hypothesis as explanation for the observed signal is less credible.
The absence of in vitro binding in FTLD-TDP might indeed be due to discrepancies in methodology as previously suggested [
22]. Firstly, numerical differences in the binding levels measured in AD compared with non-AD brains as detected by [
3H]AV1451 in vitro binding on brain homogenates seem relatively narrow [
22] when compared with the autoradiographic techniques that make use of brain sections that show a much higher signal-to-noise ratio. This may be due to the high ethanol concentrations (70-30%) used in the autoradiographic studies [
17,
41], being much more effective at washing out nonspecific binding than the PBS washes used in the homogenate binding studies [
19,
22,
24]. Moreover, in vitro autoradiography with [
18F]THK5351 (an arylquinoline) equally showed absence of binding in FTLD-TDP cases. The absence of binding therefore does not seem to be attributable to an idiosyncratic interaction between ethanol washing and [
18F]AV1451. Processing of the tissue might result in complex physicochemical interactions that may alter the secondary structure of tau fibrils and possibly, also the structure of TDP-43 aggregates. First, the use of brain sections versus brain homogenates might affect the accessibility of the tracer to the binding sites on the tau fibrils. Secondly, the use of ethanol may also affect the accessibility of the binding target due to its potential to denaturate proteins and given that ethanol is not naturally present in the brain [
51‐
53]. Therefore, we cannot exclude that [
18F]AV1451 and [
18F]THK5351 are binding to TDP-43 aggregates in vivo despite the clear absence of binding in post mortem studies.
In the SV PPA case with underlying PiD, we observed in vitro [
18F]AV1451 binding which colocalized with tau pathology and this tracer binding could be blocked by the cold compound, suggesting specific binding to FTLD tau. The binding signal was, however, four times lower than the average signal in AD. This is interesting given that in PiD, tau fibrils exist as straight filaments composed out of 3 repeat (3R) tau while [
18F]AV1451 is designed to have high affinity for the paired helical fragment structure of AD which is constituted out of 3R/4R tau [
7]. In the autoradiography study of Lowe et al. on paraffin-embedded formalin fixed tissue [
20], binding to PiD was suggested but not semi-quantified and was also lower in PiD than in AD cases. Sander et al. showed moderate [
18F]AV1451 binding on cryosections of a PiD case and displaceable binding in the 4R tauopathies CBD and PSP [
23]. These findings are in contrast with the study of Marquie et al. [
19], in which no detectable [
18F]AV1451 binding was demonstrated in cryosections containing FTLD-tau lesions from PiD (and other tauopathies: PSP and CBD) when the signal was contrasted to that of control brains. In the study of Tsai et al. [
21], a patient with sporadic bvFTD demonstrated punctate inferior temporal and hippocampus tracer retention, corresponding to the area of severe AGD pathology, a 4R tauopathy. This case also had Braak stage III [
21]. Overall, these findings, together with our current results, indicate low affinity of [
18F]AV1451 for FTLD tau.