In our sample, AoS was present in the majority (85.7%) of cases and was more often of mild (33.3%) or moderate (33.3%) severity. According to the VBM data, its severity correlated with atrophy of the SMA and SMG of the left hemisphere - structures responsible for articulation and phonological processing, respectively (Deschamps et al., 2014; Graves et al., 2023; Hertrich et al., 2016; Sliwinska et al., 2012). SMA plays a significant role in the complex planning and initiation of motor acts, including articulation, and its damage is commonly found in AoS, especially in PPAOS (Utianski et al., 2018; Josephs et al., 2012). In addition to the structural lesions, AoS also correlated with FC decline between the SMA and the putamen. These findings are consistent with a number of other studies in nfvPPA: according to Botha et al., FC between the SMA and other structures of the language network correlated with the severity of AoS in PPAOS; Carbo et al. showed that damage to white fibers between the SMA and putamen is associated with AoS, but not agrammatisms; and a study by Mandelli et al. showed that degeneration of the white matter tracts between SMA, the precentral gyrus, and putamen was one of the most significant abnormalities in nfvPPA (Botha et al., 2018; Valls Carbo et al., 2022; Mandelli et al., 2014). The association of AoS with the SMG volume was a less expected finding. It should be noted that the SMG is closely connected with the IFG pars opercularis (opIFG) and the premotor cortex and carries out integration between the phonological and articulatory components of speech (Gow, 2012; Sarubbo et al., 2016). It can be assumed that damage to the SMG may lead to a deterioration in this integration and, as a result, exacerbate articulation errors, especially in the context of existing neurodegeneration. It has also been described that lesions of the SMG can lead to other types of apraxia (Park, 2017). In addition to SMA, AoS in our study was also correlated with decreased connectivity of the precentral gyrus with both the putamen and the pMTG. The putamen has direct connections with the primary motor cortex and also participates in sequence learning and control over motor acts, including articulation (Moustafa et al., 2018; Viñas-Guasch and Wu, 2017). A decrease in this connection can lead to disruption of this regulation and be expressed in apraxia. Decreased FC between the precentral gyrus and the pMTG may serve as a reflection of disintegration between the phonological and motor components, which was also revealed in the structural GM analysis as correlations with the SMG volume. Thus, AoS in our sample was associated with damage to the structures and connections that carry out complex motor control of movements, as well as with disruption of connections between the phonological and articulatory speech components.

Along with apraxia of speech, expressive agrammatisms are the second core symptom of nfvPPA. In our sample, expressive agrammatisms were found in all patients, but in contrast to AoS, they were predominantly of minimal (32.1%) or mild (42.9%) severity. According to VBM results, its severity correlated with atrophy of the left opIFG and left insular lobe, which is consistent with the data of previous studies. The opIFG is an essential region for speech and language, and has been repeatedly associated with grammar impairment, including in nfvPPA (Whitwell et al., 2013, 2017; Tetzloff et al., 2018). The insula has extensive connections with the frontal lobe structures, including the IFG; its atrophy is characteristic to nfvPPA and was previously associated with the aphasia severity in the disease (Mandelli et al., 2016; Grossman, 2012).

Functional correlations were more widespread, involving not only frontal lobe but also the parietotemporal regions and the putamen. The identified FC disruption can be divided into several groups: FC decline between the ventral stream components (pMTG, angular gyrus, temporal pole, aITG, trIFG); between the SMG and putamen; between the premotor cortex and precentral gyrus. Damage to the ventral premotor cortex was previously described in connection with agrammatisms in the nfvPPA group and apparently reflected a decrease in speech fluency, which may explain our results (Lorca-Puls et al., 2024). Other studies have previously also shown a connection between the putamen and the SMG in grammatical disturbances in nfvPPA (Whitwell et al., 2017; Rogalski et al., 2011). However, the question remains whether their damage plays a direct role in the development of agrammatisms or affects the deterioration of grammatical function through connections with the opIFG, insula, and other areas of the left frontal lobe. Our data suggest that the putamen and SMG may play their own role in the implementation of the grammatical function of speech. Previously, coactivation of the putamen and the inferior parietal lobule, including the SMG, was shown during speech paradigms execution among healthy volunteers, another study also identified that there is a direct anatomical connection between the putamen and the SMG (Cacciola et al., 2017; Viñas-Guasch and Wu, 2017). It was also found that vascular lesions of the anterior SMG correlated with the severity of agrammatisms even after adjustment for lesion volume and speech fluency, and a small study of healthy volunteers showed the role of the SMG (along with the angular gyrus and posterior parts of the temporal lobe) in speech production when using more complex, but not simple morphosyntactic constructions (Matchin et al., 2020; Schönberger et al., 2014). Thus, decreased FC between the SMG and putamen may have a direct role in grammatical impairments in nfvPPA, but further research in this area is currently required.

Along with expressive agrammatisms, difficulties in comprehension of syntactically complex constructions is also widely present in nfvPPA. In our study, receptive agrammatisms were associated with the GM volume of frontal and temporoparietal areas of the left hemisphere, with the predominance of the latter. Early works in this field showed the importance of the posterior frontal regions, especially the IFG, in grammatical processing and that expressive and receptive agrammatisms have the same foundations (Cooke et al., 2002; Kambara et al., 2013; Friederici et al., 2006; Wilson et al., 2010a). Meanwhile, evidence is currently being accumulated on the significant role of the lateral temporal regions in this process, especially their mid-posterior parts (Matchin and Wood, 2020; Zaccarella et al., 2017). Similar data were obtained in recent studies on receptive agrammatisms in nfvPPA, where correlations were found with temporal but not frontal regions (Lorca-Puls et al., 2024; Charles et al., 2014). Our results, in turn, indicate the possible role of both regions, which has also been shown previously (Tyler et al., 2011). Lorca-Puls et al. in a recent study put forward the theory that IFG degeneration in nfvPPA can lead to the mild difficulties in sentence comprehension, whereas moderate and severe impairment develops while atrophy spreads to the left temporal lobe structures (Lorca-Puls et al., 2024). Our results partially fit this theory, but in addition indicate a significant role of the inferior parietal lobule in this process. The latter can play a role in grammatical comprehension processing by providing auditory-verbal working memory (Deschamps et al., 2014; Emch et al., 2019). For example, inferior parietal lobule coactivation with the IFG was noted in response to syntactically ambiguous stimuli, but it did not correlate directly with the complexity of the presented syntactic constructions (Tyler et al., 2011). A similar mechanism can be observed in nfvPPA, when, due to damage to the primary areas responsible for agrammatic sentence comprehension, large resources of verbal working memory are required to complete the task. In addition, as mentioned above, the SMG can also play its own role in providing grammatical functions. Our functional findings were in line with the VBM results. Receptive agrammatisms correlated with the connections between the opIFG and SMG, the anterior temporal lobe with the trIFG, the posterior temporal lobe, and the angular gyrus. The latter changes are similar to the correlations revealed in expressive agrammatisms, and may generally reflect the increased role of ventral stream regions in grammatical processing in nfvPPA in the terms of neurodegeneration.

Non-fluent speech is one of the most distinctive features of nfvPPA and can be used in differential diagnosis with other variants of the disease. Decrease in speech fluency in nfvPPA may be due to grammar and articulation impairments, but also may have its own pathogenesis. A number of studies that simultaneously assessed grammatical impairments and fluency in nfvPPA showed that they are associated with adjacent but different brain regions (Rogalski et al., 2011; Mesulam, 2016). Our work has also revealed predominantly different associations between the two symptoms with overlap in the left opIFG. While agrammatism severity correlated with the opIFG and insula volumes, decreased speech fluency was associated with atrophy of the opIFG, precentral gyrus, and superior parietal lobule (SPL). Correlations with frontal lobe regions may be explained by grammatical and articulatory impairments but the association of fluency with the SPL is less clear. Despite the fact that the direct role of this area in speech production has not been shown, the SPL is involved in maintaining attention (especially visual attention) and, according to some studies, has connections with the salience network regions, which is especially prone to neurodegeneration in frontotemporal dementia including PPA (Alahmadi, 2021; Ardila et al., 2018; Koenigs et al., 2009). At the same time, some studies indicate the influence of executive dysfunction and behavioral changes on speech rate, including in nfvPPA (Grossman et al., 2013). Thus, the association between fluency and SPL volume may be explained by the latter’s role in supporting cognitive and behavioral functions, but this requires further study. The rs-fMRI data for fluency also differ significantly from agrammatisms. Impaired speech fluency was correlated with a FC decrease of the precentral gyrus with the SMA and thalamus, as well as the putamen and thalamus thus affecting areas that provide planning and execution of motor speech. Rather similar correlations were found when assessing apraxia of speech, which emphasizes the role of the articulatory component in slowed speech rate in nfvPPA.

Impaired word comprehension is the least typical for nfvPPA and, as a rule, is a specific sign of the semantic variant of the disease (svPPA), where it is associated with the temporal pole atrophy (Botha and Josephs, 2019). In accordance with this, the symptom occurred only in a third of patients (35.7%) in our sample with predominance of minimal and mild severity (70 and 20%, respectively). According to the VBM data, single word comprehension correlated with the volume of the anterior temporal lobe of the right hemisphere. Despite the fact that semantic word knowledge is commonly associated with damage to the left temporal pole, some studies indicate the significance of bilateral damage to this area, and cases of predominantly right-sided lesions in svPPA have also been described (Binney et al., 2016; Snowden et al., 2018). Functional rs-MRI confirms the structural findings. Among the components of the left hemisphere, word comprehension impairment correlated with a FC decrease between the left temporal pole and anterior ITG with the left posterior temporal regions and left trIFG (where semantic information from the temporal pole is usually transmitted). Thus, impairments in single words comprehension in nfvPPA has structural and functional changes similar to that in svPPA, and the appearance of this symptom is most likely associated with the atrophy propagation to the temporal lobes during the course of the disease.

Repetition impairment is also not typical for the early stages of nfvPPA, and more specific for the logopenic variant of the disease (lvPPA) (Botha and Josephs, 2019). In nfvPPA repetition difficulties can be observed due to the severe apraxia of speech or dysarthria or at later stages of the disease with the development of severe aphasia or behavioral changes. In our cohort, repetition difficulties were observed in the majority of cases (89.3%), more often (80%) with very mild or mild presentation. Cases with moderate repetition impairment were observed only in patients with moderate to severe disease severity (as assessed by FRS) and aphasia duration of more than 3 years. In all cases, repetition difficulties were less pronounced than the key nfvPPA symptoms - apraxia of speech, decreased fluency and agrammatisms.

As in cases with impaired single word comprehension, our main goal in assessing repetition was to identify whether its impairment in nfvPPA has its own pathogenesis or a common basis with lvPPA. VBM showed correlations with the left SMG, which, along with the angular gyrus, is responsible for short-term phonological memory (Henry et al., 2016). Damage to these areas has been associated with the development of repetition impairment and phonological errors in lvPPA (Lukic et al., 2019; Win et al., 2017). Among functional changes, repetition was associated with FC between posterior STG with the SMG and SMA, as well as between SMA and the insula. The first finding suggests disruptions in connectivity between the dorsal stream components, connected by the arcuate fasciculus, and a deterioration in the process of transferring phonological knowledge to the posterior-frontal regions. The arcuate fasciculus connects areas of the dorsal pathway, which ensures the translation of phonological knowledge into expressive speech and its damage leads to repetition impairment and the development of conduction aphasia, which is symptomatically similar to lvPPA (Ivanova et al., 2021; Bernal and Ardila, 2009). In lvPPA, repetition impairments also correlate with the FC between the perisylvian regions and the lateral frontal and temporal regions, which was partially observed in our work (Mandelli et al., 2023). Thus, some of the correlations we identified are similar to those for lvPPA. At the same time, the correlation of repetition with the FC between the SMA and the insula indicates the presence of a motor component in repetition impairment in nfvPPA, possibly due to articulation disorders. Therefore repetition disorders in nfvPPA may be associated both with structural and functional damage to the perisylvian regions and with the involvement of the frontal regions, which suggests the presence of two possible mechanisms of their development, one of which is specific to nfvPPA and the other is similar to lvPPA.

Although naming difficulties are not characteristic for nfvPPA and more commonly occur in lvPPA or svPPA, their presence does not exclude the nfvPPA diagnosis. In our study anomia was present in more than a half patients (67.9%) with predominantly very mild (36.8%) and mild (42.1%) severity. Moderate anomia were present in four patients, all of them had disease duration >36 months and moderate to severe disease severity. In all patients naming disturbances were less pronounced than AoS and/or agrammatisms.

Structural correlates included SMG and SPL. As it was mentioned earlier, SMG is one of the areas that provide short-term phonological memory. In accordance with this, damage to SMG leads to difficulties in conversion of sound to articulation. In addition to repetition difficulties, this manifests itself in the form of anomia with intermittent speech with word-finding pauses, forgotten words can be remembered with the help of first syllable or multiple choice clue as semantic knowledge remains intact. This profile overlaps with changes found in lvPPA as naming difficulties in this PPA variant are derived mainly from violations of short-term phonological memory due to atrophy of posterior perisylvian regions (Gorno-Tempini et al., 2008). Therefore both repetition and naming in nfvPPA have some similar foundations as in lvPPA. Correlation of naming with SPL volume is more unexpected. In our work SPL volume was also associated with decreased fluency and we proposed that it could be secondary due to cognitive and/or behavioral dysfunction caused by atrophy. Similarly SPL atrophy could affect naming through its role in providing executive functions as there are some works that accentuate connection between executive dysfunction and naming disturbances (Kiss and Csépe, 2024; Higby et al., 2019).

Functional correlates were identified only after lowering statistical significance to p < 0.01. Nevertheless, the data obtained are of some interest. In contrast to the VBM findings, the association of naming with FC decrease between aITG with AG, and pITG with trIFG indicates involvement of semantic rather than phonological component of naming. pITG participates in the perception and semantic processing and transfers data from occipital lobe to the anterior parts of the temporal lobe, the main semantic hub (Hope and Price, 2016). aITG is, in turn, associated with the trIFG, and this connection plays a role in the lexical retrieval of semantic knowledge and aids in semantic processing mostly at the word level (Heide Von Der et al., 2013; Kim et al., 2022). aITG is also connected to AG, which, according to various sources, probably participates in both phonological and semantic speech processing, and has extensive connections with the temporal and frontal cortex (Junker et al., 2023; Kim et al., 2022). AG’s role in semantics remains controversial with various theories from direct role in processing semantic information to serving as a hub to accumulate semantic information and integrate it with broader cognitive processes (Zhang et al., 2024; Seghier, 2013, 2023; Graves et al., 2023; Farahibozorg et al., 2022). Regardless of the exact role, decreased connectivity between the aITG and the AG seemingly indicates violations in the semantic processing and/or integration between semantic and phonological components.

Thus, our data suggests that naming difficulties in nfvPPA may have multidirectional foundations with impaired perception, semantic and phonological processing, as well as executive impairments. At the same time, it should be noted that the fMRI data had less statistical significance and the obtained results should be treated with caution.

It should be noted that our work had several limitations. One of them was the selected scale for speech and language assessment. PASS used in our study to measure the severity of aphasia symptoms is not quantitative and allows us to divide the severity of disorders only into five groups. More precise quantification could lead to more precise results and possibly identify more correlations. At the same time, this scale was developed specifically for progressive aphasias evaluation and allows to assess all speech domains necessary for diagnosis and monitoring of patients with PPA. For these reasons and also due to absence of more accurate specialized scales for the Russian-speaking population it was chosen as the best available and suitable option. Other limitations were the relatively small sample size, which was primarily due to the low prevalence of the disease, the predominance of patients with a long disease duration (more than 3 years) and the lack of follow-up data, which did not allow us to accurately separate earlier and later neuroanatomical changes. These limitations must be taken into account when planning further research in this area.

Another relative limitation was choosing ROI-to-ROI analysis with pre-selected areas of interest for functional changes evaluation. Although selected areas covered all main known functionally significant language regions, this method cannot detect possible correlations outside the chosen regions thus artificially limiting the results. To avoid this we conducted a pilot study of whole-brain ROI-to-ROI analysis, which showed involvement of regions outside of the selected areas. However, given the small sample size and low level of statistical significance these results were not discussed in the article. Future works with more participants using seed-to-voxel or whole-brain ROI-to-ROI analysis are needed to obtain these results and expand our findings.

It is also important to note that the data obtained are among the first results for the Russian-speaking population. Thus, some differences between the data obtained and those available in the literature may be due to linguistic differences. This is especially worth considering when evaluating the results of grammar disturbances, as the most variable among different languages. At the same time, there were no native English-speaking patients in our group and we did not conduct a direct comparison between the two groups, which is why it is impossible to draw unambiguous conclusions at this stage. Further research is needed in this area with direct comparison between native speakers of different languages to identify possible differences.