Introduction:
Perhaps the most fascinating part of the human brain is the dorsolateral prefrontal cortex [1]. It holds much that we value for cognition, such as working memory, spatial manipulation, visualization and executive function. But despite its immense importance, it tends to be an under-researched area.
One reason for this is simply the difficulty of studying it. Rodents lack both the dlPFC and rostral PFC areas (like the frontal pole [2]), so the subject species required are humans and primates. Cajal noted a precursor of this a hundred years ago, while mice have short, sparingly branched basal dendrites, primates have many long, heavily branched dendrites, with the branching increasing with "psychic" functioning. Combined with a focus on studying the hippocampus, this renders the dlPFC under-studied. Importantly, it also renders a negative result in rodents insignificant, because the region in question simply doesn't exist there.
The dlPFC is also relatively recently evolved, bordering the frontal pole cortex (Brodmann areas 10 and 10p [3]). And unique among most cortical regions, it depends far more on NMDAR than AMPAR activation. Its delay-period firing is more than five times as NR2B-dependent [4] as the equivalent firing in the visual cortex. This alone can explain much of the brain's vulnerability to aging in this area.
Delay and Response Cells:
The dlPFC allows primates to hold information in memory without anything external to refer to, or in essence, creating and maintaining a representation with no direct sensory input, the fundamental mechanism of abstract thought [5]. As Goldman-Rakic suggested, this single operation is the basis of every other higher function, like behavioral control and cognitive control, and much else besides.
Delay cells perform this task. Found within the pyramidal microcircuitry in deep layer III (the "psychic" layer), these neurons are intensely recurrently wired, which allows them to fire throughout the working memory delay period [6]. Their responses are spatially tuned, so they respond vigorously to their preferred orientation and only weakly, or not at all, to other orientations.
The reader should distinguish Delay cells from Response cells, which operate very differently. During a working memory task, Response cells are generally inhibited during the delay period and instead fire around the time of the motor response, preparing or executing an action, or providing feedback on one. Unlike Delay cells, they reside mainly in layer V, where they are selectively influenced by D2, and where the bulk of the mRNA for all five dopamine receptor subtypes is found. That receptor distribution [7] is not uniform: D1R and D2R are more concentrated in the striatum than in cortex, D4R and D5R are higher in cortex, and D3R is roughly equal across both. D1R in particular localizes to the spines, and produces the familiar inverted-U on Delay cell firing, where too much or too little dopamine causes the representation to collapse.
View attachment 58520
The persistent firing of the Delay cells relies entirely on NMDA receptor activation. This is demonstrated with iontophoresis: a brief exposure to the AMPAR antagonist CNQX barely affects the cells, whereas the NR2B antagonist Ro25-6981 [8] completely abolishes their firing. Further, the NR2B-selective antagonist TCN-237 destroys the cells' ability to code visual space. This suggests the firing depends not just on NMDA activation generally, but on the specific NR2B subtype.
The dependence of dlPFC function on catecholamines is astonishing. After depleting dopamine in a discrete dlPFC region [9], monkeys in a spatial-delayed-alternation task showed nearly as much of a deficit as in an ablative lesion study, and could be successfully treated with apomorphine or L-DOPA. Removal of dopamine from the dlPFC caused almost as great a behavioral deficit as total removal of the cortex.
View attachment 58521
The lesion studies also reveal a left/right asymmetry in dlPFC function: damage disrupts spatial manipulation and verbal memory separately [10], with left-hemisphere damage selectively impairing working memory and right-hemisphere damage impairing broader reasoning. The region also seems to play a role in threat detection, consciousness [11], pattern detection, risk-taking, and affect and disease, namely ADHD and schizophrenia. We can also infer some involvement in physical endurance [12].
Dynamic Network Connectivity:
The binding mechanism tying all the previous regions together is what Arnsten's group has termed Dynamic Network Connectivity [13] (DNC). It relies on the dynamic alteration of the strength of connections in the layer III recurrent microcircuit. Rather than the synaptic remodeling of LTP/LTD seen in the hippocampus, DNC rapidly and reversibly alters the strength of existing connections without altering physical spine structure. The layer III dlPFC spines stay remarkably similar despite age, they remain long and thin even in old monkeys, where hippocampal spines would have remodeled, and that thin morphology is precisely what lets a single ion channel near the synapse gate the connection on or off quickly. It is also what makes this area both the most adaptable and the most fragile, since a DNC-based control mechanism gives great flexibility, but under improper neurotransmitter conditions it becomes labile. By tuning catecholamine levels it integrates arousal and cognitive function at the same time, and an unusual number of schizophrenia genes were found to have some function linked to DNC.
View attachment 58523
The next compound mentioned, guanfacine, is a good example of what happens here: it binds the α2A-adrenergic receptors, reducing intracellular cAMP. This closes the potassium channels, strengthening connectivity.
View attachment 58524
The Top-Down Dichotomy:
The most recently evolved structures of the prefrontal cortex are located rostrally and laterally; these areas appear to control their older counterparts medially and caudally. The dlPFC communicates with the amygdala directly through Brodmann area 24 [14], and indirectly to Brodmann area 25 [15]via BA10m and BA32. It sits at the top of the control hierarchy for the executive and attention systems, and disruption by lesion, stress or electrical stimulation results in disordered behavior [16].
View attachment 58525
We can also now see exactly how this fails. The cingulate cortex activates the amygdala, and too much catecholamine in the higher cortical areas effectively takes the dlPFC offline [17] for the task at hand, this is DNC running in reverse, where feedforward Ca²⁺-PKC and cAMP-PKA signaling opens K⁺ channels and disconnects the network. Increased dopamine activity and strong stimulation [18] may trigger the same mechanism. So you get into a cycle with no brakes, the primal circuits running unchecked, the basolateral amygdala's stress-induced hyper-excitability feeding into motivated behavior and addiction. This seems more direct for women, where emotion-versus-working-memory leads to a significantly larger amygdala and OFC response in women than in men [19].
View attachment 58526
View attachment 58527
A lot of antidepressant approaches, like serotonergic ones and psychedelics, work at least partly by inhibiting BA25, thereby getting the cortex back in the game to take over control. rTMS and insight therapies seem to work through the opposite mechanism, directly enhancing the top-down driving force from the left dlPFC.
This is really what is so compelling about the dlPFC-amygdala relationship and its potential to be the brain's gatekeeper. You can make a good plan from within the cortex, like initiating a habit or executing on the things that are actually meaningful to you, and then the basal ganglia and amygdala override you on emotionality or effort cost, and you don't do it. And the reward for strengthening the dlPFC side shines through repeatedly. Excitatory neuromodulation of the dlPFC gives a durable reduction of craving and consumption [20] in addiction and overeating, and a single course of active dlPFC tDCS significantly improved executive function and reduced craving immediately after and one month later, while placebo had no effect [21].
View attachment 58528
Guanfacine is the proof of concept, since it increases Delay cell firing via α2A agonism and makes the dlPFC-amygdala relationship more cognitively controlled [22], although its side-effect profile keeps the dose limit small. The dlPFC is also a huge player in motivation and anticipation [23], feeding reward signals into the mesolimbic and mesocortical DA pathways [24].
If amygdala overactivity reduces cognitive control, then keeping top-down control, using your intelligence instead of returning to immediate gratification, may well be about the most desirable goal we can aim for, neurologically and philosophically. But inhibiting BA25 is only a partial approach, and the real leverage is provided by the mGluR2/3 system, where a single mechanism has the biggest impact on dlPFC firing rates.
mGluR2 and mGluR3 Pull in Opposite Directions
This is the area with the best, most primary evidence, and it comes straight from Jin, Arnsten et al. 2017 [25] rather than from summary. Both mGluR2 and mGluR3 are Group II mGluRs, historically viewed as presynaptic autoreceptors [26] that inhibit glutamate release [27], but that isn't true in the dlPFC. They are both present postsynaptically as well, with layer III containing mostly presynaptic mGluR2 and mostly postsynaptic mGluR3 located within the spine synapses.
View attachment 58529
This is confirmed by the clinical pattern, where declining cognitive function in both schizophrenia and aging is correlated with reduced mGluR3, whereas losing mGluR2 seems somewhat neuroprotective in these contexts. This only makes sense if the receptors are antagonistic toward each other rather than working in a similar fashion, given their historic role as Group II autoreceptors. mGluR3 appears to increase Delay cell firing while mGluR2 reduces it, and mGluR2 appears to be the more effective at suppressing glutamate release.
So it makes sense to simultaneously up-regulate mGluR3 and inhibit mGluR2. These effects appear to balance out each other's side effects, with the combined action being more effective than treatment at either receptor individually. There are only a few known compounds that attempt to achieve this simultaneously, like PHCCC(4Me) [28] (also known as THCCC) and the more desirable CID 49858305 [29], but these are highly novel, with limited safety predictions and essentially no published data on their effects at the moment (related chemistry and background here) [30].
Increasing mGluR3: NAAG and GCP-II:
This is an opportune moment to bring up NAAG and GCP-II. NAAG is an endogenous neurotransmitter that promotes mGluR3 activity [31]. While it is still debated whether NAAG is a full agonist, there are clues. In one study, purified NAAG appeared non-agonistic [32], suggesting it might act as a positive allosteric modulator and that earlier reported agonism was due to glutamate contamination, and other evidence found no Group II or NMDA interaction at all [33]. Proper study on pure NAAG is still needed, but there is clearly some form of pro-mGluR3 signaling activity.
GCP-II is a glial enzyme that metabolizes NAAG, so inhibiting GCP-II raises endogenous NAAG and enhances mGluR3 signaling. Administering NAAG directly seems to have the undesirable effect of causing unpreferred firing, which suggests targeting the enzyme is a better way to reach optimal NAAG levels. Increased NAAG also correlates with superior cognitive function in both rodents and primates; carrying the rs202676 polymorphism [34] is linked with increased GCP-II, lower NAAG, and lower IQ in humans, consistent with reduced mGluR3 signaling. Since the dlPFC sits atop pain pathways, it is likely that GCP-II inhibition will also be analgesic, with the possibility of a long-lasting reduction in pain [35].
Looking at the actual compounds available for mGluR3, ZJ-43 [36] seems the best candidate currently. Its advantages are improved working memory across a variety of animal models [37], positive (grain-of-salt) predictions, better BBB penetration than 2-PMPA, and the likelihood that nasal-spray delivery increases absorption. The drawback is the requirement for high doses, likely 75-100mg intranasally or 100-150mg orally.
The slightly more potent GCP-II inhibitor is 2-PMPA [38]. It can roughly double Delay cell firing [39], but suffers from limited BBB penetration and some off-target activity at cytosolic carboxypeptidases [40], specifically Nna1. It is important to note this is a structure-dependent effect of 2-PMPA, not something inherent to GCP-II inhibition in general. A separate, earlier compound, 2-MPPA, proved well tolerated and showed no cognitive impairment even at extremely high doses in humans [41], but had to be withdrawn over structure-related concerns.
It is worth noting that while rodents given chronic 2-PMPA saw reduced extracellular 5-HT in the dorsal striatum [42], this may owe to the rodents' minimal dlPFC; humans who received the related 2-MPPA did not see this effect even over several days, and no such result has ever been observed in human subjects. If you remain concerned, mGluR2 antagonism should push 5-HT and 5-HT2A in the desired direction anyway, and that antagonism tends to come bundled with the agents we want.
Reducing mGluR2:
As mentioned earlier, mGluR2 is the more potent glutamate suppressor [43]. Inhibiting it has been proposed as one of the mechanisms behind the antidepressant effects of psychedelics [44]. On cognitive function, mGluR2 (not mGluR3) exhibits an inverted dose-response curve [45] for both working memory and Delay cell firing, yet there appears to be no cognitive impairment reported after the use of any mGluR2 NAM so far. Inhibiting mGluR2 actually seems to be one of the downstream cognitive effects of MR agonism [46]; and the MR agonists themselves are useful cognitive enhancers [47] but cannot be applied clinically due to their side effects.
In combination, mGluR3 enhancement coupled with mGluR2 inhibition should precisely tune the glutamate system rather than broadly inhibit it
References:
- https://en.wikipedia.org/wiki/Dorsolateral_prefrontal_cortex
- https://pubmed.ncbi.nlm.nih.gov/21388858/
- https://en.wikipedia.org/wiki/Brodmann_area_10
- https://pmc.ncbi.nlm.nih.gov/articles/PMC6041807/
- https://www.cell.com/neuron/pdf/0896-6273(95)90304-6.pdf
- https://journals.physiology.org/doi/abs/10.1152/jn.1989.61.2.331
- https://pmc.ncbi.nlm.nih.gov/articles/PMC4485014/
- https://pmc.ncbi.nlm.nih.gov/articles/PMC3584418/
- https://pubmed.ncbi.nlm.nih.gov/112679/
- https://pubmed.ncbi.nlm.nih.gov/22789779/
- https://pubmed.ncbi.nlm.nih.gov/35231420/
- https://pubmed.ncbi.nlm.nih.gov/18514741/
- https://pubmed.ncbi.nlm.nih.gov/37020275/
- https://pubmed.ncbi.nlm.nih.gov/20554470/
- https://pubmed.ncbi.nlm.nih.gov/4998921/
- https://pubmed.ncbi.nlm.nih.gov/6165607/
- https://pubmed.ncbi.nlm.nih.gov/30878500/
- https://pmc.ncbi.nlm.nih.gov/articles/PMC2907136/
- https://pmc.ncbi.nlm.nih.gov/articles/PMC5611728/
- https://pubmed.ncbi.nlm.nih.gov/17544015/
- https://pubmed.ncbi.nlm.nih.gov/34514666/
- https://pubmed.ncbi.nlm.nih.gov/32289681/
- https://pubmed.ncbi.nlm.nih.gov/25059532/
- https://pubmed.ncbi.nlm.nih.gov/37035621/
- https://pmc.ncbi.nlm.nih.gov/articles/PMC5974790/
- https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1460-9568.2004.03378.x
- https://www.frontiersin.org/journals/cellular-neuroscience/articles/10.3389/fncel.2018.00508/full
- https://pubchem.ncbi.nlm.nih.gov/compound/PHCCC_4Me
- https://pubchem.ncbi.nlm.nih.gov/compound/49858305
- https://pubs.acs.org/doi/10.1021/jm101069m
- https://www.sciencedirect.com/science/article/abs/pii/S1074742719300061
- https://pubmed.ncbi.nlm.nih.gov/19389924/
- https://pubmed.ncbi.nlm.nih.gov/19285517/
- https://psychiatryonline.org/doi/10.1176/appi.ajp.2020.19111152
- https://pubmed.ncbi.nlm.nih.gov/23776202/
- https://pubchem.ncbi.nlm.nih.gov/compound/11722553
- https://pmc.ncbi.nlm.nih.gov/articles/PMC6397785/
- https://pubchem.ncbi.nlm.nih.gov/compound/10130754
- https://pubmed.ncbi.nlm.nih.gov/35732693/
- https://doi.org/10.1016/j.bbrc.2020.10.029
- https://bpspubs.onlinelibrary.wiley.com/doi/full/10.1111/j.1365-2125.2005.02396.x
- https://core.ac.uk/download/pdf/154342699.pdf
- https://pmc.ncbi.nlm.nih.gov/articles/PMC3690373/
- https://pmc.ncbi.nlm.nih.gov/articles/PMC3064746/
- https://pubmed.ncbi.nlm.nih.gov/28108498/
- https://pubmed.ncbi.nlm.nih.gov/25178162/
- https://www.sciencedirect.com/science/article/abs/pii/S1074742716302519?via=ihub
