Evaluation of in vitro neuronal platforms as surrogates for in vivo whole brain systems

Quantitatively benchmarking similarities and differences between the in vivo central nervous system and in vitro neuronal cultures can qualify discrepancies in functional responses and establish the utility of in vitro platforms. In this work, extracellular electrophysiology responses of cortical neurons in awake, freely-moving animals were compared to in vitro cultures of dissociated cortical neurons. After exposure to two well-characterized drugs, atropine and ketamine, a number of key points were observed: (1) significant differences in spontaneous firing activity for in vivo and in vitro systems, (2) similar response trends in single-unit spiking activity after exposure to atropine, and (3) greater sensitivity to the effects of ketamine in vitro. While in vitro cultures of dissociated cortical neurons may be appropriate for many types of pharmacological studies, we demonstrate that for some drugs, such as ketamine, this system may not fully capture the responses observed in vivo. Understanding the functionality associated with neuronal cultures will enhance the relevance of electrophysiology data sets and more accurately frame their conclusions. Comparing in vivo and in vitro rodent systems will provide the critical framework necessary for developing and interpreting in vitro systems using human cells that strive to more closely recapitulate human in vivo function and response.


Results
Spontaneous single-unit activity. Neuronal electrophysiology was compared between rodent cortical neurons cultured on planar MEAs (in vitro) and awake, freely moving rats using implantable MEA probes (in vivo). Key similarities and differences for single-unit activity between these two neuronal systems were assessed by comparing both baseline electrophysiological activity and response to two classes of well-characterized neuroactive chemicals. Spontaneous (baseline) single-unit extracellular neuronal spiking activity in both in vitro and in vivo systems was compared by quantifying features typically observed for both systems [24][25][26][27][28] . Most notably, in vitro firing events were typically in bursts, whereas in vivo firing events were predominantly composed of single spikes (Fig. 1a,b; Supplementary Fig. S1). We also noted greater variability within the recorded features among experimental replicates (firing rate, interspike interval, and reduced single spike activity outside of bursts) in the in vitro system (Fig. 1c-e).
Drug effects on spontaneous single-unit activity. Upon exposure to atropine, single-unit responses for firing rate and non-burst activity were similar between the systems ( Supplementary Fig. S2a,b). Importantly, the expected dual response was observed in both neuronal systems, whereby isolated single units exhibiting simultaneously increasing or decreasing responses to atropine were observed both in vitro (Fig. 2a,b) and in vivo (Fig. 2c,d) after exposure to 1 µM atropine. This dose of atropine caused no statistically significant differences >60% in vitro), although the overall change in firing rate for these cells does not change with dose (Fig. 2f). For both concentrations evaluated, only the 5 µM exposure demonstrated statistically significant differences in the single-spike activity between the two systems ( Fig. 2g). Cells in vitro appeared more likely to increase in firing rate after atropine exposure, although the response was not statistically significant compared to the in vivo response from similar exposures (Fig. 2i). Changes in single-spike activity also showed a similar but statistically insignificant trend (Fig. 2j).
Next, we investigated the effects of ketamine in both systems. Studies have reported that at sub-anesthetic doses in vivo, ketamine induces both increases and decreases in cortical neuron firing rate. At higher, anesthetic doses, most neurons exhibit a reduced firing rate, while a small but significant subset of neurons still exhibits a firing rate increase 20 . In the current study, we observed ketamine's dual response behavior both in vitro and in vivo (Fig. 3). Overall, cortical neurons in vitro demonstrated greater sensitivity to ketamine exposure, with approximately one order of magnitude decrease in the dose range in vitro required to elicit the same response as in vivo (Supplementary Fig. S3). The difference in sensitivity is further illustrated in Fig. 3e and Supplementary  Fig. S2c,d, which examines single-unit responses for the two concentrations of ketamine (45,90 µM) that elicited responses in both systems. Nearly all cells in vitro (>99%) showed a decrease in activity when exposed to either 45 or 90 µM ketamine. In vitro, this decrease is reflected as a near complete cessation of activity (Fig. 3f); this was not the case for ≤1 µm ketamine ( Supplementary Fig. S3) where a subset of in vitro cells showed increased activity. In vivo, however, fewer cells demonstrated decreases in activity for both 45 µM and 90 µM ketamine (68.8% and 37.8%, respectively). This reduction corresponded to a 40-50% decrease in activity for these cells (Fig. 3f). Interspike intervals (ISI) for both systems did not show a significant change from baseline (0 µM ketamine, Fig. 3h), aside from the in vitro exposure to 90 µM ketamine, which showed a complete (but reversible, Supplementary Fig. S4) cessation in single-unit activity. Finally, Fig. 3g shows a trend towards less non-burst firing in vitro compared to in vivo, similar to the changes in firing rate seen in Fig. 3f. This does not indicate an increase in bursting, but instead indicates decreases in single spike activity overall when interpreted in the context of the firing rate decreases in Fig. 3f.

Discussion
This study aimed to quantify the similarities and differences of cultured cortical neurons and cortical neurons in vivo by comparing spontaneous electrophysiological activity and changes in electrophysiology after exposure to atropine or ketamine. These experiments showed: (1) significant differences in spontaneous in vivo and in vitro firing activity, (2) similar response trends in both systems in single-unit spiking activity after exposure to atropine, (3) greater sensitivity to the effects of ketamine in neurons in vitro. It is important to be aware that several factors could influence the differences observed in this study. These factors include: the age of neurons (embryonic versus adult) 29 , the ratio of GABA to glutamatergic cells 15 , presence and composition of supporting glial cells 30 , and the isolated environment for the in vitro system which lacks input from other organs 31,32 . Each of these points will be discussed in greater detail below.
In vitro cultures typically rely on embryonic neurons, as cell viability after harvesting and long-term culturing is more robust in embryonic and postnatal rodent systems 33,34 compared to adult [35][36][37] . Electrical properties for neurons cultured in vitro at different ages have been compared and found to be similar across age groups 29 apart from bursting behavior, which was more common in embryonic than adult neurons 29 . This characteristic of embryonic neurons may contribute to the differences observed here for spontaneous baseline neuronal activity between the two systems. Cultured neurons demonstrated more burst-like features (i.e. fewer spikes out of bursts) compared to neurons in vivo. With these inherent differences between systems in mind, responses to chemical stimuli were evaluated relative to the spontaneous (baseline) activity of the given system. This evaluation allowed for a more direct comparison for responses between systems.
The composition and ratios of neuronal subspecies, as well as the level of their direct interaction and feedback ability, are considerably different between in vivo and in vitro systems. Previous work has shown that a NMDA antagonist similar to ketamine, MK-801, decreases the neuronal activity of pyramidal cortical cells in culture and anesthetized rats 38 while increasing cell activity in pyramidal cortical cells in awake rats 39 . This difference was attributed to decreased GABAergic interneuron activity in the awake animal, resulting in disinhibition of cortical neurons and a subsequent increase in spiking activity 27 . In cultured neurons or in anesthetized neurons in vivo, this disinhibition of cortical neurons is ineffective due to low basal levels of extracellular glutamate, the additional factor required to induce pyramidal cell action potentials 40,41 . Additionally, rodent cortical neurons in culture show an increased sensitivity to the excitotoxic effects of glutamate 42 and neuronal NMDA receptors show differing sensitivity depending on the developmental stage of the neurons, with a hypersensitive stage seen in the first two weeks of rat neonatal life 43 . These factors, as well as the differences in GABAergic neurons between systems, may have contributed to the differential responses observed for ketamine. Within our in vitro system a small percentage (~8%, Supplementary Fig. S5) of neurons are GABAergic, compared to ~20% in vivo 15 . Supporting glia composition also differed between systems. Neuroglia including astrocytes, oligodendrocytes, and microglia all play a role in maintaining and influencing neuronal health and function 44,45 . While our in vitro system contains some contaminating astrocytes (GFAP + , ~13%, Supplementary Fig. S5), biologically relevant ratios and other cell types are not present. Cortical regions in vivo contain a much larger percentage of glial support cells (~40%) 46 which are known to help protect against deleterious glutamate effects and possess NMDA receptors themselves 47 .
While differences in composition and ratios of cell types between systems may have influenced the response to ketamine, atropine responses were more comparable. Expression of muscarinic receptors is low in the embryonic rat cerebral cortex (E18; 5% of adult levels) but receptor expression increases rapidly through the first 11 days postnatally 30 before stabilizing at levels that will be maintained through adulthood 48 . Interestingly, Heacock et al. showed that despite low expression of muscarinic receptors in embryonic neurons, their sensitivity to the agonist carbachol was similar to that of adult neurons 30 . The authors attributed the comparable sensitivity to a change in effective turnover rate (or the addition of unused receptors to cerebral cortical neurons as rats age) to explain why the neurons respond similarly to muscarinic agonists despite the differences in the maturity of the cultured neurons. A similar argument could explain why our in vivo and in vitro systems responded similarly to atropine. Additionally, muscarinic receptor expression on supporting cell types such as microglia and oligodendrocytes does not change with age in the rodent model 49,50 . Therefore, the recognized differences in ratios of neuronal to supporting cells for both systems may have not affected overall response.
As isolated tissue mimetics, in vitro model systems lack the ability to metabolize drug compounds. Therefore, the metabolites found after drug administration in vivo, and their potential off-target effects, are likely not recapitulated in vitro. This represents a significant difference between the two systems, particularly the differing responses seen for ketamine, a drug known to be readily metabolized in vivo. Importantly, ketamine metabolites have known neuroactive properties, and can act on other receptors to counter ketamine effects 21 or directly work on the same receptors as ketamine with less efficacy 31 . These neuroactive metabolites may have contributed to the differences observed between our experimental systems.
Neuromodulatory inputs from other regions are absent in the in vitro system; these inputs influence cortical neurons in vivo 32 . The cortex is ~25% of the rodent brain's total volume 51 but is comprised of many specialized subregions that are a result of these interregional inputs. The input for each individual cell in the cortex is influenced by the larger circuitry of the brain; this can also have additional, complex effects on the ultimate response a single neuron shows to systemically applied drugs.
Taken together, our results demonstrate that there are quantifiable differences and similarities in extracellular electrophysiology between in vivo and in vitro cortical systems both for spontaneous activity and in response to chemical challenge with atropine and ketamine. This study represents only a preliminary step towards a more thorough comparison between in vivo and in vitro responses, which would require testing additional representative drug compounds across multiple drug classification categories. These more extensive comparisons are necessary to correlate and benchmark the responses between the two systems to better interpret in vitro data in the context of the whole organism. While the complexity of the in vivo system is a key determinant of chemical response, directly demonstrating the ability of the relatively simple in vitro system to recapitulate the fundamental response observed in vivo is critically important. Using two different classes of drugs, we demonstrate the utility of cortical in vitro neuronal systems to recapitulate whole organ responses to certain classes of chemicals. While we show that changes in extracellular electrical activity in neuronal cultures may be appropriate for certain types of pharmacological studies, we also demonstrate that this system may not fully recapitulate the responses observed in vivo. Understanding the boundaries of a representative response with this in vitro system will provide more relevant data sets, and a framework to develop in vitro systems that more closely mimic in vivo function and response. As such, in addition to expanding the number and type of drug compounds to test, future work should be aimed at developing more complex cultures (i.e. additional cell types, three dimensional cultures) that capture more of the whole brain response for complex drugs of action like ketamine, while concomitantly verifying that functional responses pivot towards a more in vivo functional phenotype. However, until more complex cultures or alternative in vitro systems have been developed and validated, we hope to begin a larger conversation between in vitro and in vivo communities to understand the right questions we can correctly answer with in vitro platforms to directly contribute to human health and disease.

Methods
Chemicals. All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless otherwise noted.
Ketamine HCl (100 mg/mL in sterile water) was purchased from ACE surgical (Brockton, MA).
Microelectrode device fabrication. The MEA device used for in vitro platform testing was microfabricated as previously described 52,53 . Briefly, MEAs (20-50 µm platinum electrodes spread evenly over 1.8 mm 2 ) were fabricated on a glass substrate. Standard photolithography and wet or plasma etching were used to pattern metal constituting the 60 electrodes and electrode traces; polyimide was deposited and patterned as an insulating layer over the metal traces. A polystyrene cylinder was affixed over the MEA to enable cell culturing over the array. The well area was 113 mm 2 and accommodated approximately 700 μL. ZIF connectors were added to the device for electrical connections and electrodes were plated with platinum black 53 . Impedance measurements were taken prior to seeding and ranged from ~50-250 kΩ at 1 kHz.
The MEA devices implanted in vivo were microfabricated and plated as described previously 54,55 . Briefly, the implantable MEAs featured two separate shanks (each 15 μm thick and 100 µm wide) spaced 2 mm apart medial-laterally. At the tip of each shank, 16  (dorsal-ventrally through medial prefrontal cortex). MEA devices were fabricated on silicon substrates using alternating stacks of metal (conductive layers) and polyimide (insulation layers). Standard photolithography and wet or plasma etching were used to pattern the 32 electrodes, trace metal, and device outlines. Polyimide was deposited and patterned, followed by deposition of platinum for the electrode metal. The electrode metal area was then patterned and etched creating electrodes 20 μm in diameter, followed by an insulating polyimide layer. Next, layers of gold (for the metal traces connecting the electrodes to the connector region) and polyimide were deposited to create a layered structure. After fabrication, devices were released from the silicon wafer and individually connected to 36-channel Nano Strip connectors (Omnetics Connector Corporation, Minneapolis, MN) using conductive epoxy. Electrodes were then plated as described above with final impedances ranging from 50-400 kΩ at 1 kHz and prepped for surgical implantation.
In Vivo Surgical Implantation. All animal experiments and surgical procedures were conducted following the guidelines and regulations set by Lawrence Livermore National Laboratory, including Institutional Animal Care and Use Committee approval (#240). Devices were surgically implanted into medial Prefrontal Cortex (2.52 mm anterior-posterior, 1.2-1.8 mm medial-lateral, 1.5 mm dorsal-ventral) of male Sprague-Dawley rats (6-8 weeks, ~200 g). The device, with a silica insertion tool temporarily attached to the probe with polyethylene glycol (PEG), was implanted using a microdrive positioner (Kopf Instruments, Tujunga, CA). The PEG holding the positioner was then dissolved with saline and the insertion tool was retracted with the microdrive, leaving the array implanted in the brain 54,56 . The array connector was then affixed to the skull with dental acrylic.
Chemical dosing. For in vitro administration, chemicals were either dissolved (atropine; powder, >99.9% purity) or diluted (ketamine) in primary neuron basal media (PNBM) to 4X the final desired in vitro concentration. Cultures used for chemical testing were between day in vitro (DIV) 19 and 78. For chemical exposures, 25% of the culture media was removed from the MEA well and replaced with an equal volume of 4X chemical working stock (0.0004-360 µM ketamine, 4-10 µM atropine), resulting in a final culture concentration that ranges from 0.0001-90 µM for ketamine and 1-5 µM for atropine.
For in vivo administration, chemicals were either dissolved (atropine powder) or diluted (ketamine solution) in 0.2-0.4 mL saline (0.9% NaCl in sterile water). Intraperitoneal (i.p) dosing concentrations were based on published studies that evaluated brain region specific accumulation of non-metabolized (radiolabeled) drug upon i.p. injection 57,58 . Atropine (10 mg/kg) was given i.p. to rats (400-550 g); this was equivalent to 1 µM atropine in cortex (and 50 mg/kg = 5. For in vitro electrophysiology measurements, devices were placed on a heated stage (37 °C) in a 5% CO 2 controlled chamber during recordings. Electrophysiology measurements were acquired using a multi-channel recording system (AlphaLab SnR, Alpha Omega, Alpharetta, GA) and were sampled at a frequency of 22.3 kHz and bandpass filtered between 268 and 8036 Hz. Baseline and chemical exposure measurements were recorded for 20 minutes. Chemical exposure measurements were collected for 20 minutes immediately following chemical addition. After testing, cultures were given at least two days for recovery between experiments.
For in vivo measurements, drug application was randomized so that there were no more than two recordings from the same animal in the same week. Each rat (N = 3) was administered both concentrations of each drug three times over a period of 6-12 weeks with at least two days between experimental drug exposures. For each drug administration, animals were placed in an operant behavioral chamber (Med Associates, Inc., St. Albans, VT) and acclimated to the chamber for >15 minutes before recording was initiated. Electrophysiology was recorded using a multichannel freely moving animal system (Tucker-Davis Technologies, Alachua, FL) at 24,414.1 Hz with a band pass filter of 100-5000 Hz. Baseline was monitored to ensure proper connectivity of devices and noise thresholds (auto set to ±4.0 standard deviations above noise). Animals then were administered a vehicle injection of 0.2-0.4 mL saline (0.9% NaCl in sterile water, i.p.) and 20 min of baseline condition data was collected. Next, animals were injected with the chemical challenge dissolved in 0.2-0.4 mL saline (0.9% NaCl in sterile water, i.p.) and >20 min of chemical condition data were then collected. Each single unit was analyzed individually and considered a data point for the experiment since each cell could respond differently to both drugs; averaging responses across a device would confound results and interpretation. Electrophysiological data from each unit on the in vitro MEA device or in vivo probe was analyzed using parameters we have previously evaluated on cortical and hippocampal cultures seeded on MEA devices 53 . Bursts were defined as having a maximum beginning interspike interval of 0.1 sec, a maximum end interspike interval of 0.2 sec, a minimum interburst interval of 0.5 sec, a minimum burst duration of 0.05 sec, and a minimum of 10 spikes per burst 53 . Feature analysis was carried out with an in-house custom R package.
Calculated features were inspired by those described in the work of Charlesworth et al. 24 : interspike interval (ISI), overall firing rate and percent of spikes outside of bursts. For drug challenges, the difference in electrophysiological activity before and after challenge was calculated and reported as the change in mean frequency from baseline. Here, n indicates the number of isolated single units. Single units that showed a complete cessation of firing activity with drug challenge were defined at a -100% change in activity. Single units exhibiting greater than 1000% change in the feature were excluded. Data was considered non-normally distributed for statistical analysis and were analyzed using GraphPad Prism7. In vivo and in vitro baseline firing were compared with the non-parametric Mann-Whitney test (Fig. 1c-e). For drug challenges, a chi-square test was run on the proportions of single units showing an increase, decrease, or no change for both systems (Figs 2e and 3e). Differences between multiple groups for features (Figs 2f-j and 3f-h, Supplementary Figs S2 and S3) were evaluated by the non-parametric Kruskal-Wallis one-way analysis of variance (ANOVA), followed by Dunn's post hoc test. A p value of <0.05 was considered significant. All error bars indicate standard error of the mean (SEM).
Code availability. The analysis code used to analyze data in this study is available upon request.
Histology. In vivo probe placement was verified after CO 2 euthanasia of animals. Rats were perfused using a 10% formalin solution. Brains were removed, cryoprotected, and stored in 10% formalin for >3 days. Tissue was then sectioned coronally into 40-50 μm thick slices on a cryostat (Leica CM3050 S, Buffalo Grove, IL). The probe tract was identified visually under a microscope equipped with a camera. Approximate recording sites are shown in Supplementary Fig. 6. Data availability. The data that support the findings of this study are available from the corresponding authors upon reasonable request.