Distinct Lateral Prefrontal Regions Are Organized in an Anterior-Posterior Functional Gradient

Abstract

The dorsolateral prefrontal cortex (dlPFC) is composed of multiple anatomically defined regions involved in higher-order cognitive processes, including working memory and selective attention. It is organized in an anterior-posterior global gradient where posterior regions track changes in the environment, whereas anterior regions support abstract neural representations. However, it remains unknown if such a global gradient results from a smooth gradient that spans regions or an emergent property arising from functionally distinct regions, that is, an areal gradient. Here, we recorded single neurons in the dlPFC of nonhuman primates trained to perform a memory-guided saccade task with an interfering distractor and analyzed their physiological properties along the anterior-posterior axis. We found that these physiological properties were best described by an areal gradient. Further, population analyses revealed that there is a distributed representation of spatial information across the dlPFC. Our results validate the functional boundaries between anatomically defined dlPFC regions and highlight the distributed nature of computations underlying working memory across the dlPFC.SIGNIFICANCE STATEMENT Activity of frontal lobe regions is known to possess an anterior-posterior functional gradient. However, it is not known whether this gradient is the result of individual brain regions organized in a gradient (like a staircase), or a smooth gradient that spans regions (like a slide). Analysis of physiological properties of individual neurons in the primate frontal regions suggest that individual regions are organized as a gradient, rather than a smooth gradient. At the population level, working memory was more prominent in posterior regions, although it was also present in anterior regions. This is consistent with the functional segregation of brain regions that is also observed in other systems (i.e., the visual system).

Keywords: anterior–posterior; dlPFC; gradient; parcellation; prefrontal; working memory.

Figure 1.

A global functional gradient can result from smooth or areal gradients along the anterior–posterior axis of the dlPFC. A, Schematic of the macaque brain, highlighting dorsal and ventral dorsolateral prefrontal regions (top left, inset) with the posterior tip of the principal sulcus marked with a white circle (right). A.S., Arcuate sulcus; P.S., principal sulcus. B, A smooth gradient of functional properties in the dorsal dlPFC, where a one-segment linear model (black line) fits the data better than a multisegment discontinuous piecewise model (gray lines). C, An areal gradient of functional properties in the dorsal dlPFC, where a multisegment discontinuous piecewise model (black lines) fits the data better than a one-segment linear model (gray line). Vertical dashed lines represent estimated anatomic boundaries between areas 8Ad, 9/46d, and 46d. Zero on the x-axis denotes the posterior tip of the principal sulcus.

Figure 2.

Overview of methods. A, The monkeys performed a delayed-saccade task with distractor interference. The periods analyzed are shown in the colored horizontal bars, including target (red), distractor (green), and delay (yellow). B, We recorded from the dorsolateral prefrontal cortex of two macaque monkeys (Macaca fascicularis) and estimated the anterior–posterior and dorsoventral electrode positions from surgery images (data not shown). C, We estimated native space anatomic parcellations using a neuroimaging atlas of M. fascicularis; D, Electrode positions in one monkey (M2) and anatomic parcellation boundaries (black dashed lines). E, To test whether the global gradient is the result of a smooth or an areal gradient across a range of functional measures, we tested whether a one-segment or a two-segment discontinuous piecewise model better described the data, separately for the dorsal and ventral dlPFC and separately for each monkey. Top, The adjusted R² of the one- or two-segment model fit. (Significant fits are highlighted in bold font and with an asterisk.) Regression lines are shown for the one-segment model (left) and two-segment model (right). The white vertical dotted line represents the estimated functional boundary between 8Ad and 9/46d for this measure. F, From the previous analysis, we derived functional parcellation boundaries (black dotted line) and grouped neurons based on this functional parcellation map constructed for each monkey. G, Using the functional parcellations in F, we assessed population-level functional differences between regions using cross-temporal decoding analyses and measuring target information quantity (decoding performance) and stability across regions (neurons pooled across both monkeys).

Figure 3.

Analysis of anterior–posterior gradients in the dlPFC across PCs derived from physiological measures. A–E, We fit smooth one-segment and areal two-segment models spanning areas 8Ad and 9/46d across five PCs, which have a cumulative explained variance of ≥75% (loadings for each measure in columns III and VI). Left to right, The columns of plots depict the smooth and areal model of two monkeys (M1, I–II; M2, IV–V). Model fits that are significant versus a shuffled baseline are plotted in full color saturation and denoted with an adjusted R² (*p < 0.05; n.s., p > 0.05). For each monkey, we compared the model fits of the one- and two-segment models versus the alternative model with AIC, where ΔAICc = 0 represents the best fit of the pair (columns I vs II, III vs VI). Order of measures is as follows: white (stimulus presentation period measures), target selectivity F statistic, distractor selectivity F statistic, response latency, receptive field size, stimulus selectivity index, distractor filtering; black (delay period measures), delay 1 selectivity F statistic, delay 2 selectivity F statistic, nonlinear mixed-selective F statistic, delay 1 memory field size, delay 2 memory field size, delay 1 selectivity index, delay 2 selectivity index.

Figure 4.

Decoding of the target location in correct trials. A, Cross-temporal decoding of four target locations for both dlPFC subregions. B, Information quantity during target presentation period, quantified as the performance along the diagonal of the target presentation period. C, Information quantity during delay periods 1 and 2 (average of both delays, denoted by boxes D1 and D2), quantified as the performance along the diagonal of delays 1 and 2. D, Within-period code stability during the target presentation period, quantified as the ratio of the average performance during target presentation period (without the diagonal), to the diagonal E, Within-period code stability during the delay periods, quantified as the ratio of the performance during the delay period (D1 and D2, without the diagonal), to the diagonal (average of both delays), F, Across-delay code stability, quantified as the ratio of decoding performance of the decoder trained in delay 1 and tested in delay 1 versus delay 2 (vice versa for delay 2).