Multiple component networks support working memory in prefrontal cortex

Abstract

Lateral prefrontal cortex (PFC) is regarded as the hub of the brain's working memory (WM) system, but it remains unclear whether WM is supported by a single distributed network or multiple specialized network components in this region. To investigate this problem, we recorded from neurons in PFC while monkeys made delayed eye movements guided by memory or vision. We show that neuronal responses during these tasks map to three anatomically specific modes of persistent activity. The first two modes encode early and late forms of information storage, whereas the third mode encodes response preparation. Neurons that reflect these modes are concentrated at different anatomical locations in PFC and exhibit distinct patterns of coordinated firing rates and spike timing during WM, consistent with distinct networks. These findings support multiple component models of WM and consequently predict distinct failures that could contribute to neurologic dysfunction.working memory | prefrontal cortex | macaque | coherence H igh-level cognition depends on the ability to translate stored information about recent experience into a behaviorally appropriate response, an ability known as working memory (WM). WM relies on a storage process that actively maintains information and a control process that manipulates stored information to support the selection and preparation of a contingent response (1-3). The neural mechanisms that support WM involve networks that are broadly distributed throughout the brain (4-7) and rely heavily on the prefrontal cortex (PFC) for normal operation (6-9). However, the degree to which WM is supported by a single distributed network or multiple specialized network components in PFC remains unclear (6, 10, 11), hindering progress in the search for neurocognitive therapies to treat disorders of cognition (12).Persistent spiking activity is commonly thought to reflect the mechanistic basis of WM in PFC (13-16). This activity manifests in different ways, including time-varying neuronal responses that decay, ramp up, or are stable in time during memory delays. Although such a diversity of responses could reflect distinct modes of persistent activity, it has long been a standard practice to treat all persistently active neurons in PFC as representative of a single composite WM function that supports the maintenance and manipulation of information necessary for memory-guided behavior (14,(17)(18)(19). The implicit assumption that the representations of stored information and contingent responses overlap at the neural circuit level contrasts with an alternate view, which suggests that PFC primarily encodes the selection and preparation of responses (6,10,11). This difference highlights the need to directly investigate the circuit-level organization of storage and response preparationrelated activity in PFC.We address this problem here, using a simple manipulation of WM in concert with large-scale recordings from neurons across lateral PFC of macaque monkeys. By mapping neural activity during memory and visual del...

Correlated variability modifies working memory fidelity in primate prefrontal neuronal ensembles

“…A previous experiment (30) reported this trend but did not find a significant effect, perhaps due to a smaller sample size (295 pairs, compared with our 10,535 pairs). A second prediction is that WM maintenance should modify r sc as a function of r signal ; r sc should be lower between neurons with dissimilar tuning than neurons with similar tuning (18,(29)(30)(31)(32)(33). Indeed, we found that the relationship between r sc and r signal changed as predicted during the delay, compared with the fixation and stimulus epochs (Fig.…”

“…2A) (P < 0.005 for all epochs, bootstrap test) (Materials and Methods); and r sc varied as a function of tuning similarity, which we quantified as signal correlation between pairs of neurons during the delay epoch (r signal ) (Materials and Methods) (Fig. 2B) (18,(29)(30)(31)(32)(33). Specifically, we found that the median pairwise r sc was consistently larger for similarly tuned neuron pairs (defined as r signal > 0.25) compared with dissimilarly tuned pairs (defined as r signal < −0.25) (P < 0.001 for all epochs, bootstrap test) (Materials and Methods).…”

“…Reports of the effects of ensemble r sc structure on information content vary significantly in sign and magnitude (19,22,24,27,42,48,49); our results can help reconcile these disparate accounts. For example, previous studies that applied decoding techniques to simultaneously recorded ensembles found that removing the r sc structure decreased decoding accuracy for grating orientation (48) and remembered location (49) whereas another study reported a positive effect of pairwise r sc on information coding (32). Moreover, spatial attention increases signal-to-noise primarily by reducing r sc (25)(26)(27).…”

“…These results are difficult to extend to WM coding in the LPFC. Furthermore, there are relatively few studies investigating r sc in the LPFC (21,27,(29)(30)(31)(32)(33); and only one of these studies directly examined the effects of r sc on information coding (27). Prior results examining pairwise correlations are also difficult to extrapolate to larger neuronal ensembles, which have a complex, multidimensional r sc structure that cannot be characterized by pairwise measurements alone (20).…”

Neural Manifolds for the Control of Movement

“…A number of studies have shown that the largely heterogeneous activity patterns of individual neurons in monkey (Kobak et al, 2016; Machens et al, 2010; Mante et al, 2013; Markowitz et al, 2015) and rat (Durstewitz et al, 2010) prefrontal cortex, monkey (Churchland et al, 2010b) and rat (Forsberg et al, 2016) V1, rat olfactory cortex (Kobak et al, 2016), rat thalamus (Chapin and Nicolelis, 1999), rat parietal cortex (Raposo et al, 2014), locust olfactory system (Stopfer et al, 2003), aplysia pedal ganglion (Bruno et al, 2015), and perhaps the entire zebrafish brain (Ahrens et al, 2012) can be explained as generated by a small set of latent variables associated with neural modes. In all these studies, neural modes and their time-varying activation helped describe previously unexplained mechanisms of neural function.…”