A Cool Approach to Probing Speech Cortex

Abstract

In this issue of Neuron, Long et al. (2016) employ a novel technique of intraoperative cortical cooling in humans during speech production. They demonstrate that cooling Broca’s area interferes with speech timing but not speech quality.

A critical issue neurosurgeons face is delineating the extent of cortical tissue that can be safely resected. If you remove too little tissue, one may not get the desired clinical results. On the other hand, if you remove too much tissue you risk impairing motor and language function postoperatively. In the 1930s, Wilder Penfield and colleagues applied electrical stimulation directly to motor and somatosensory cortex during neurosurgical procedures. Their findings provided a systematic mapping of motor function and culminated in the famous cortical homunculus diagram (Penfield and Boldrey, 1937). Over the next 20 years, Penfield devised protocols for awake language mapping (motor cortex can be mapped under anesthesia) that remain largely unchanged to this day (Penfield and Roberts, 1959).

In addition to providing an invaluable tool for neurosurgeons, electrical stimulation mapping (ESM) is a unique technique for probing function in a causal manner. During a typical language mapping procedure, the patient is awaken from anesthesia while cortex is still exposed and is asked to perform the same tasks Penfield first employed over 80 years ago: counting and picture naming. While the patient engages in counting or naming, different cortical sites are repeatedly stimulated. If the applied electrical current reliably disrupts the patient’s speech output, or causes naming errors with intact speech output, the cortical site is deemed to be critical for language and is spared from resection. Even though counting and picture naming tasks tap into a relatively small subset of language functions, sparing of sites identified by ESM dramatically reduces postoperative deficits and the procedure remains the gold standard in the field (Chang et al., 2015; Sanai et al., 2008).

A large current is necessary (the actual threshold varies by patient but is on the order of 10 mA) to achieve a reliable behavioral deficit during stimulation. Applying such a stimulating current repeatedly can cause neuronal after discharges, which can develop into a seizure in either diseased or healthy tissue. Historically, intravenous pharmacological agents were used to halt stimulation-evoked seizures but their efficacy has a temporal delay and since these agents dampen cortical activity in both the epileptic and healthy brain, this treatment interferes with subsequent stimulation mapping. To circumvent these issues, a new approach was developed by Sartorius and Berger which rapidly quenches seizure activity by applying cold Ringer’s solution (or saline in some cases) to the exposed cortex. Notably, this cooling method does not interfere with neuronal excitability when normal temperature is restored (Sartorius and Berger, 1998), permitting continued mapping of eloquent language cortex.

Cortical cooling techniques have been extensively used to produce localized and reversible inactivation of neurons in a variety of animal models and multiple sensory systems (Brooks, 1983). By employing cryogenic probes, a limited surface of cortex can be cooled to a desired temperature (the effect of cooling is confined to 1.5–2.5 mm) (Lomber et al., 1999). Large temperature gradients (reduction from normal cortical temperatures to below 20°C) are used to locally suppress neuronal activity and after termination of cooling the tissue rapidly recovers function both in animals (Lomber et al., 1999) and in humans (Bakken et al., 2003). Recently, findings in the songbird provide evidence that applying modest cooling to premotor HVC slows neuronal spiking, rather than completely abolishing it, and causes the bird to sing slower. In this issue of Neuron, Long et al. (2016) extend this technique to humans by employing intraoperative cooling probes to regions of Broca’s area and precentral gyrus in order to investigate their causal role in speech production.

In their study, neurosurgical patients undergoing awake language mapping were recruited. Similarly to Penfield’s original patients, ESM was first performed to identify critical language sites for clinical use. These sites were then used to guide cooling probe locations. Regions of Broca’s area and precentral gyrus were cooled (Figures 1B and 1C) while patients produced overlearned sequences of speech (the days of the week, e.g., Monday–Friday, or a string of numbers, e.g., 21–25). Unlike electrical stimulation, cooling of cortex infrequently caused speech arrest but rather caused a degradation of speech output that manifested in either the quality of the speech utterance (dysarthric, slurred speech) or in the timing of the speech output (compressed or expanded utterances). Long et al. (2016) quantified the quality of speech using raters on Amazon Mechanical Turk who each rated speech segments from one patient on a scale from 0 (“Extremely degraded”) to 1 (“Typical/Normal”). They found that the degradation was site specific and co-varied with cortical temperature. As an important control, a subset of patients performed an unrelated motor task in which they pressed a button as fast as they could. The same sites that caused speech degradation did not have an effect on button press behavior, indicating that these sites are speech specific or at least mainly related to mouth-motor articulators.

Figure 1. Converging Evidence from Cooling and Electrophysiology.

(A–C) Cortical sites that showed speech timing or quality effects (A) when a cooling probe (B) was placed on pars triangularis (PTri; mostly effecting timing, yellow), pars opercularis (POp; mostly effecting timing, yellow), and precentral gyrus (PrCG; mostly effecting quality, aqua) (C). Adapted from Long et al. (2016).

(D and E) Sites of electrophysiological recordings showing peak activity prior to (blue) and during (red) word production (D). Note that pre-speech sites are evident in pars triangularis, pars opercularis, superior temporal gyrus, and superior temporal sulcus (all blue) and articulation sites (red) center on precentral gyrus. Vertically stacked single trial traces of high-frequency activity across temporal and frontal cortices showing cessation of activity in Broca’s areas prior to articulation (black line; E). Adapted from Flinker et al. (2015).

Long et al. (2016) then investigated whether cooling prefrontal sites led to changes in speech timing, akin to the birdsong results. They identified spectrotemporal elements of the patients’ recorded speech (i.e., entire words, a gap between words, a segment of a word) and compared the length of these elements under cooling condition with normal speech (i.e., the same element of speech when no cooling was applied). While a majority of cooling sites showed no effect on speech timing, a quarter of the sites showed an expansion of speech (slowing) and around 15% showed an unexpected compression of speech (acceleration). Both the speech timing and quality measures were then projected on to a normalized brain in order to assess the relative contribution of cortical regions. Cooling locations in the motor cortex (precentral gyrus) resulted in many more quality changes than timing changes. Conversely, cooling of Broca’s area (pars opercularis and triangularis) resulted in more timing changes than quality changes (Figure 1A). This dissociation is quite striking in face of the partial covariance of both measures (i.e., slower speech could be rated as more degraded). It would be interesting to further analyze how speech degradation manifests (e.g., articulation, phonation, resonance, and prosody) and ascertain to what degree the errors arise from degraded motor planning versus degraded articulatory coordination.

These findings complement a growing body of evidence that is redefining how we view the role of Broca’s area in speech production. Traditionally, Broca’s area has been viewed as a center for speech output. Nevertheless, lesion studies have shown that cortical damage limited to Broca’s area does not cause a Broca’s aphasia but rather results in a transient, improving mutism (Mohr et al., 1978). Likewise, speech arrest during intraoperative stimulation mapping is much more probable in the ventral precentral gyrus than in Broca’s area (Tate et al., 2014). Recently, electrophysiological intraoperative recordings from two separate groups have shown that activity in Broca’s area precedes speech output by ~250 ms (Flinker et al., 2015; Magrassi et al., 2015) and that Broca’s area is involved in forming an articulatory plan rather than coordinating the articulators, which are supported by motor cortex (Figures 1D and 1E) (Flinker et al., 2015). These findings fit nicely with the data provided by Long et al. (2016). Cooling of motor cortex disrupts coordination of the articulators and causes degradation of speech quality, while cooling of Broca’s area influences the articulatory plan causing temporal changes in speech but not degradation of speech quality (Figure 1A).

Unlike electrical stimulation, cortical cooling does not spread as much across cortex and does not interfere with electrophysiological recordings. Future studies could leverage combined recordings of neuronal activity while different components of the speech network are perturbed. Long et al. (2016) propose an interesting hypothesis that Broca’s area contains a site responsible for sequence generation in speech production. In order to fully investigate this hypothesis, it will be crucial to test more complex speech sequences as well as employ stimuli that are not overlearned and that fully engage the linguistic circuit. In summary, Long et al. (2016) contribute an invaluable “cool” tool to our arsenal in human neuroscience and open a new exciting chapter in speech research.