Word comprehension in temporal cortex and Wernicke area

Methods

Results

All 4 regions showed significant cortical thinning and volume loss in the patient group. Mean volume loss, computed in comparison to volumes in the controls, was highest in AT (28.4%), least in IFG-B (9.6%), and intermediate in MT (20.3%) and PT-W (14.5%) (figure 1A).

Discussion

The variability of the boundaries proposed for the Wernicke area and the controversies surrounding its relevance to language comprehension have prompted recent calls to eliminate the usage of this eponym altogether.³⁴ However, the term Wernicke area is so entrenched in the literature that such recommendations, while rational, are unlikely to be adopted. The goal of the current study was not to rediscover or redefine the location of the Wernicke area, an enterprise that may be of great historical interest but not particularly productive for exploring the neurobiology of language. Instead, we aimed to investigate the differential sensitivity of word comprehension to cortical damage along a rostrocaudal axis of temporal neocortex, an axis that incorporates association pathways linking unimodal percepts, such as word forms, to corresponding transmodal associations that mediate their recognition as concepts.³⁰ The posterior part of this axis, PT-W of figure 1, represents a composite of arguably all sites that have been proposed over the years as components of the Wernicke area.

In the group of 73 patients with PPA who participated in this study, word comprehension impairment showed no significant correlation with cortical thinning (i.e., volume loss) within the parietotemporal cortices that encompass even the most expansive delineation of the Wernicke area (PT-W in figure 1). A second analysis by individual vertex rather than entire ROI showed that the absence of correlation with comprehension characterized atrophy not only in PT-W as a whole but also in its individual components such as pSTG, SG, AG, and pMTG (figure 2). Furthermore, the postmortem examination of a patient with PPA showed that severe neuronal degeneration in this region may leave word comprehension essentially intact. The strongest anatomical correlates of word comprehension were located anterior to PT-W, particularly within the rostral third of the temporal lobe. These results are in keeping with most prior investigations on PPA and semantic dementia but in conflict with classic principles of aphasiology and some of its current reformulations.

Our failure to implicate the Wernicke area in word comprehension could be questioned by pointing out that residual neurons might have sustained performance on the PPVT. Although regions of even severe atrophy may still contain neurons that appear intact,²⁰ 2 considerations can be offered to counter this objection. First, PT-W atrophy was correlated with impairment of repetition. It is difficult to see why residual neurons of PT-W would selectively support comprehension but not repetition. Second, in the 1 patient where the cellular basis of PT-W atrophy could be quantitated by microscopic analysis, severe neuronal degeneration in the pSTG and SG components of the Wernicke area impaired repetition but not word comprehension. Although the null hypothesis is always difficult to prove, this evidence strongly suggests that word comprehension is not particularly sensitive to the integrity of PT-W and, therefore, of the Wernicke area.

The PT-W sector in this study included the posterior third of the STG but not the primary auditory cortex of Heschl gyrus. According to DeWitt and Rauschecker's³⁵ meta-analysis, the STG contains an anteriorly directed synaptic hierarchy that links auditory input first to a region selectively tuned to phonemes and then to an area specialized for transforming phonemes into auditory word forms. The absence of correlation between auditory word comprehension and PT-W atrophy suggests that the pivotal word form areas are located more anteriorly, possibly within the MT sector, and that the input they receive from upstream auditory areas, originating in Heschl gyri of both hemispheres, remains preserved.

Repetition of phrases and sentences was the 1 function most sensitive to cortical abnormalities of PT-W, and specifically its posterior STG component. This relationship is in keeping with investigations on stroke-induced conduction aphasia and logopenic PPA, showing that the parietotemporal junction is crucial for the integrity of phonologic encoding and auditory working memory.^5,6,10,36,37 Regression analyses had previously shown that impairments in the comprehension of syntactically complex sentences are correlated with cortical thinning in the IFG as well as the SG and AG components of PT-W.³ Since the current investigation did not find a correlation between sentence comprehension impairment and volume loss in PT-W, compromise of this function may be contingent on injury not only within the Wernicke area but also within frontal components of the language network.

The large-scale network model posits that each network node mediates critical (or essential) as well as ancillary (or sustaining) functions related to its principal cognitive domain.^30,38 While initial stages of damage to a given node may not cause fixed impairments of its ancillary functionalities, the overall computational flexibility of the network for mediating that task may be compromised. Such disruptions in domains of ancillary functionality can be detected by tasks that impose heavy computational loads. These principles apply to the role of the Wernicke area in language comprehension. For example, patients with agrammatic and logopenic PPA whose atrophy tends to be concentrated in PT-W and who have normal PPVT scores display abnormally prolonged semantic interference effects and loss of the N400 semantic incongruence potential.^39,40 Furthermore, fMRI investigations using synonym identification tasks revealed activations not only in the AT sector but also in posterior temporal cortex.^41,42 It appears, therefore, that parietotemporal cortex where the Wernicke area is located has an ancillary but not essential role in single word comprehension. In fact, single case studies show that an intact Wernicke area cannot sustain word comprehension in patients with AT atrophy.⁴³ The decline of BNT scores on the second but not initial examination of our PPA patient who came to autopsy (figure 3) also suggests that extensive atrophy in the Wernicke area does eventually interfere with the lexicosemantic task of object naming. However, object naming appears less sensitive to neurodegeneration in the Wernicke area than repetition and also more dependent on the integrity of AT cortices.

We found a strong correlation between atrophy of the AT and impairment on the PPVT. This correlation suggests that word comprehension is an essential function of AT cortex, a conclusion that is consistent with previous work on PPA, semantic dementia, and herpes simplex encephalitis.^{14,15,43–45} The critical role for word comprehension attributed to the AT sector is also consistent with more recent reports based on multimodal functional mapping of language and principal component analyses in patients with stroke.^10,35 This relationship was neither detected nor suspected by classic aphasiology, probably because AT is not susceptible to focal cerebrovascular lesions. The area we delineated as AT is heterogeneous and includes the temporal pole as well as the anterior components of STG, MTG, and ITG. It should be pointed out that the anterior fusiform and parahippocampal gyri, which may also play important roles in comprehension,^43,46 were not included in any of the sectors involved in these analyses. Although the results in this report are confined to the left hemisphere, our previous work had shown that unilateral left AT lesions are sufficient to cause severe word comprehension impairments so that the association of this region with comprehension impairment is not necessarily related to its known propensity for bilateral degeneration.⁴³

Object naming and word comprehension impairments were also associated with atrophy in the mid-sectors of the temporal lobe (MT in figure 1). This association deserves a cautious interpretation. MT is rarely, if ever, the initial site of peak atrophy in PPA. Atrophy spreads into this region either from AT (in semantic PPA) or from PT-W (in logopenic PPA).³² Even when MT is a site of peak atrophy, word comprehension remains unaffected if AT is not also atrophied.³ We therefore conducted additional analyses and found that the MT relationships with naming and comprehension remained just as strong when the effect of PT-W atrophy was factored out but disappeared when the effect of AT atrophy was factored out. The correlation of MT atrophy with word comprehension and naming impairments may therefore be contingent on the extension of neurodegeneration from AT in the course of disease progression.

We have seen patients where AT atrophy was associated with severe deficits of object naming in the absence of additional word comprehension impairments. In some of these patients, word comprehension impairments emerged when the peak atrophy spread posteriorly into the MT.³ It is therefore conceivable that word comprehension is most consistently impaired when AT and MT are both damaged. In some cases, AT atrophy alone seems sufficient to undermine word comprehension but we have not seen cases where isolated MT atrophy (or anterior extension of atrophy from PT-W to MT) has been associated with such deficits. This formulation helps to explain why left AT lobectomy can lead to naming but not necessarily comprehension impairment as lobectomies rarely extend into MT.⁴⁷ Several recent models of the language network have emphasized the key role of the MTG for word comprehension.^{6,11,12,48,49} The critical part of the MTG identified in many of these models is located posteriorly and falls within the PT-W sector of figure 1. In others, it does extend more anteriorly into the MT sector of figure 1. Our results suggest that mid-MTG damage influences word comprehension but that this may become clinically consequential only in the presence of additional destruction (or perhaps disconnection) of cortices within the AT sector.

Nearly all contemporary textbooks of neurology, more than a century of classic aphasiology, and several current lesion-symptom mapping studies state that language comprehension is linked to the integrity of parietotemporal regions where the Wernicke area is located. These statements cannot be dismissed lightly and need to be reconciled with our findings. An important fact to consider is that the linkage of parietotemporal cortex to comprehension is almost always based on patients with cerebrovascular rather than degenerative disease. Furthermore, the lesions tend to be large and include most of the PT-W territory rather than being confined to the pSTG or SG sectors at the core of the Wernicke area. These considerations lead to 2 models of reconciliation: one based on disconnection and the other on diaschisis (figure 5).

The reconciliation through disconnection rests on the fact that vascular territories include not only cortex but also deep white matter tracts. In the context of a stroke encompassing the PT-W sector, injury to the cortex of the Wernicke area would disrupt repetition, as predicted by the correlations in figure 2. Moreover, the additional destruction of deep white matter tracts coming from otherwise intact posterior and contralateral cortices would trigger a broader deafferentation of the language network, and might cause the word and sentence comprehension impairments characteristic of Wernicke aphasia. In contrast, degenerative injury to the cortex of the Wernicke area as seen in PPA, while also impairing repetition, would cause only a partial disconnection of the anterior language areas, insufficient to undermine comprehension. The clinical outcome would be a logopenic aphasia. Accordingly, the cortex of the Wernicke area is not all that important for word comprehension, whereas the Wernicke region is pivotal because it also lies at the confluence of deep white matter tracts, the exact identity of which remain to be determined. In keeping with this formulation, Wernicke aphasia is almost never reported in PPA, and is not part of its current classification system.¹⁹

The second potential reconciliation relies on diaschisis, which denotes a severe postsynaptic depression of neuronal activity that transcends the effect of simple disconnection. Diaschisis can be transient, as in spinal shock, or permanent, as in the case of crossed cerebellar atrophy. A stroke in the Wernicke area could elicit word comprehension impairments through diaschisis in the AT sector, explaining why the most consistent relationship of the Wernicke area to Wernicke aphasia has been reported in the acute stages of cerebrovascular accidents. In fact, clinical experience shows that stroke-induced Wernicke aphasia frequently resolves into conduction aphasia in the chronic stage as diaschisis is resolved and as the cortical dysfunction of the Wernicke area becomes the defining substrate of the aphasia. In neurodegenerative disease, however, the indolent pace of neuronal injury precludes diaschisis and therefore word comprehension remains relatively preserved.

It is also worth considering the possibility that the indolent pace of neuronal loss in PPA may have triggered synaptic reorganization elsewhere in the network. Functions normally mediated by the Wernicke area could therefore have shifted to other cortical regions. This possibility cannot be ruled out but is unlikely for 2 reasons. First, it would be difficult to envisage a mechanism of reorganization that would spare word comprehension but not repetition upon destruction of neurons in the Wernicke area. Second, the phenomenon would have to be unique to the Wernicke area since a similar process of compensatory reorganization does not seem to relieve the word comprehension impairment of AT atrophy or the agrammatism of IFG atrophy. Nonetheless, the question of synaptic reorganization in neurodegenerative diseases deserves further investigation. Such processes have been demonstrated in PPA but whether they represent aberrant rewiring or compensatory reorganization remains to be resolved.^42,50

Quantitative atrophy–deficit correlations in PPA are converging toward the view that the cortical neurons of the Wernicke area (defined broadly as the PT-W sector or narrowly as pSTG) play an ancillary rather than critical role in word comprehension and that the principal contribution of this area in the language domain is to mediate phonologic encoding and repetition. The function of word comprehension, once linked to the Wernicke cortex, is moving to more anterior parts of the temporal lobe. This historic shift in the organization of the language network can potentially be reconciled with classic aphasiology by considering the physiologic and hodologic differences between the effects of stroke and those of neurodegeneration.

Author contributions

M.-M. Mesulam: study concept and design. B.M. Rader: acquisition and summaries of data. J. Sridhar: acquisition of data, preparation of images in figures. M.J. Nelson: critical revision of the manuscript and intellectual content. J. Hyun: critical revision of the manuscript and intellectual content. A. Rademaker: data analysis and critical revision of the manuscript and intellectual content. C. Geula: neuropathologic images and data summary, critical revision of the manuscript and intellectual content. E.H. Bigio: neuropathologic images and data summary, critical revision of the manuscript and intellectual content. C.K. Thompson: critical revision of the manuscript and intellectual content. T.D. Gefen: neuropathologic images and data summary, critical revision of the manuscript and intellectual content. S. Weintraub: critical revision of the manuscript and intellectual content. E. Rogalski: image analysis and critical revision of the manuscript and intellectual content.

Study funding

R01 DCOO8552 from the National Institute of Deafness and Communication Disorders and P30 AG013854 from the National Institute on Aging.

Disclosure

The authors report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.

Publication history

Received by Neurology June 13, 2018. Accepted in final form September 19, 2018.