Sustained and potent analgesia with negligible side effects enabled by adaptive individualized granular stimulation in rat brainstem

Objectives. To clarify if an adaptive current stimulation protocol, in which current amplitude is modulated during continuous stimulation, provides better efficacy than constant current stimulation protocol with respect to analgesia caused by individualized stimulation in rat periaqueductal gray matter (PAG) /dorsal raphe nuclei (DRN). Approach. Ultrathin microelectrodes adapted for recording (n = 6) and stimulation (n = 16) were implanted in rat primary somatosensory cortex and PAG/DRN, respectively. In each animal included (n = 12), a subset of PAG/DRN microelectrodes (n = 1–3 per animal) was selected that on simultaneous stimulation blocked nociceptive withdrawal reflexes in awake unrestrained animals without noticeable side effects. Analgesic effects were subsequently assessed from both nociceptive withdrawal reflexes and intracortical pain-related responses on CO2 laser hind paw stimulation. The analgesic effects of adaptive current PAG/DRN stimulation comprising incremental increases of 5 μA/microelectrode (initial median current 30 μA/microelectrode) when effects declined were compared to the effects of constant current stimulation. Behavioral effects and brain state related changes were analyzed using quantitative movement analysis and electrocorticography (recorded on top of the dura mater), respectively. Tissue reactions and probe placement in PAG/DRN were assessed with immunohistochemistry. Main results. Powerful and sustained (4 h) analgesia was achieved with the adaptive current protocol within a rather wide area of PAG/DRN. Analgesic after-effects were seen for up to 30 min. Behavioral and brain state related side effects were minimal. Moreover, 6 weeks after implantation, there were no traces of bleedings, only small glial reactions and small but not statistically significant loss of neurons nearby indicating that the microelectrode stimulation employed is biocompatible. Significance. The results indicate that sustained and powerful analgesia with minimal side effects can be achieved by granular and individualized stimulation in PAG/DRN using an adaptive current stimulation protocol. This microelectrode technology and stimulation paradigm thus has the potential of providing a highly efficient and safe pain therapy.


Introduction
The management of chronic pain is still a major medical and societal challenge since available treatments are often both insufficient and associated with severe adverse effects. Since the 60s, it is known that powerful analgesia can be elicited by stimulation in several different sites in the brain [1][2][3][4][5][6]. One such analgesic center is in the PAG/DRN, a complex midbrain structure activated during life threatening situations [7][8][9]. The complex organization, both with regard to local circuitry and input-output connections [10], may partly explain why stimulation-produced analgesia has often been accompanied by various side effects (such as escape or aversive reactions, vocalization, rotations, and freezing) [7,10,11].
Clinically, DBS of this region, based on conventional electrodes in the millimeter scale, can provide strong analgesia [12][13][14]. Unfortunately, stimulationproduced intolerable side effects are difficult to avoid due to the spread of stimulation current to unwanted sites [15][16][17][18][19]. Another problem in brain stimulation with conventional electrodes in the clinic is that the analgesic efficacy often decays over time periods such as days and even months [12,20,21]. For short time scales, a loss of efficacy may depend on phenomena such as adaptation [22][23][24][25][26], synaptic plasticity in nociceptive pathway and, phenomena related to fatigue. For longer time scales, tissue reactions, encapsulation of implanted electrodes or even damage to the tissue may also play a role [27][28][29][30][31][32][33].
The above-mentioned problems may, at least partly, be mitigated by implanting multiple microelectrodes combined with individualized selection of efficient subsets of microelectrodes and exclusion of microelectrodes that do not provide therapeutic effects and/or induce side effects [34]. Using this technique in the PAG/DRN region, a powerful and very selective analgesia has been reported in animal studies, manifested as essentially abolished pain-related cortical responses, as well as abolished nociceptive motor responses, without noticeable behavioral side-effects [34]. A remaining problem is that the analgesic efficacy often decays over time. This decay may depend on changes in therapeutic thresholds in the stimulated area, synaptic plasticity in nociceptive pathways, phenomena related to fatigue, tissue reactions, encapsulation of implanted electrodes or even damage to the tissue induced by the stimulation [12,21,35,36]. Together the shortcomings have restricted the use of brain stimulation in PAG/DRN for clinical treatment of chronic pain.
The main objective of the present study was to clarify if an adaptive current stimulation protocol for microelectrode-based stimulation methods, employing stimulation of individually selected sites in rat PAG/ DRN, can provide prolonged analgesia. To further assess the clinical potential of this technique, stimulation-induced effects on normal behavior (documented by video tracking of animal movements), brain state (ECoG) as well as tissue responses to the implanted microelectrodes in PAG/DRN (immunofluorescence microscopy) were analyzed.
Pain-related intracortical recordings (from six implanted microelectrodes in S1) and nociceptive withdrawal reflexes were analyzed to assess pain in awake freely moving rats. It is known that nociceptive-evoked potentials recorded from S1 in humans strongly correlate to perceived pain magnitude [37,38]. These potentials have similar features to the ones recorded from animals [39,40], thus providing a translational model of pain. In the same animals, individualized stimulation in PAG/DRN (subset of 1-3 microelectrodes selected from a cluster of 16 implanted microelectrodes) was made. ECoG recordings were made using another electrode implanted on top of the dura mater.
The results show that a powerful and long-lasting analgesia can indeed be obtained by an adaptive current stimulation protocol for individualized stimulation in PAG/DRN for at least 4 h. No behavioral sideeffects, very small brain state related ECoG effects and only mild tissue reactions to the implanted microelectrodes were seen. Taken together, these results indicate that individualized granular stimulation in PAG/DRN has the potential for providing highly valuable pain therapies in the clinic.

Animals and implantation
A total of 13 Female Sprague Dawley rats was used. They were kept in a 12 h light/dark cycle at 21 • C and 65% humidity with food and water as needed. The rats were anesthetized with 2% isofluorane (kept between 1% and 2% for the rest of the surgery; Isobavet, Apoteksbolaget, Sweden) in40% oxygen and 60% nitrous oxide and then fixated in a stereotactic frame (Neurostar, Robot Stereotaxic instrument, Germany) equipped with a heating table kept at 36.6 • C. The skull was exposed and cleaned with saline solution, four anchoring screws (Agnots, MCS1x2, Sweden) were inserted, and three skull holes were drilled (Kopf model 1474 High-Speed Stereotaxic Drill, USA) for probe implantations under a stereomicroscope. Target coordinates for stimulation and recording microelectrodes were set in a Neurostar robot stereotaxic instrument (Neurostar, StereoDrive 4.0.0, Germany) to: 7.8 mm caudal, 0 mm lateral and 6.2 mm ventral to bregma [34] and 1.5 mm caudal, 2.4 mm lateral (left hemisphere), 1.45 mm ventral to bregma [44], respectively. Two insertion speeds were used: 1000 µm s −1 till 2 mm from target and 100 µm s −1 to target with 3 min pause in between for stimulation electrodes; 1000 µm s −1 till 450 µm from target and 100 µm s −1 to target without pause for recording electrode. An ECoG electrode, L-shaped at the tip, was placed on top of the dura mater a few millimeters frontal to bregma. All probes were fixated with dental cement (RelyX Unicem Self-Adhesive Universal Resin Cement) to the skull. Before awakening, the rats were injected subcutaneously with 0.01 mg kg −1 of (Temgesic Buprenorfi, Schering-Plough, Belgium) for post-operative pain relief and with 5 ml of saline to avoid dehydration. For all the experiments including animals, the anim-als´care was in accordance to Malmö/Lund Animal Ethics Committee on Animal Experiments guidelines (ethical permit M4480-18).

Nociceptive stimulation and cortical recordings
CO 2 laser stimulation of the plantar side of the right hind paw (at least 3 s in between each stimulus; randomized skin sites) was used to selectively activate nociceptive afferents (figure 1(A)) [45,46]. Nociceptive evoked cortical responses and nociceptive withdrawal reflexes on CO 2 laser stimulation (wavelength, 10.6 µm; 5 W, unfocused beam, 3 mm in diameter, MedArt VariMed Diode Laser System, Denmark) were used to assess pain in the animal [34]. It has been shown that the late evoked potentials and MUA in S1 cortex (e.g. figure 4(A), delimited by the IOI, see section 2.6.2) are not dependent on non-painful input from tactile fibers driven by reflex activity [40]. These cortical evoked potentials can be elicited also during anesthesia in the absence of withdrawal reflexes [34]. Moreover, granular PAG/DRN stimulation, of the type used here, can selectively abolish the nociceptive responses in S1 cortex while having marginal effects on responses to tactile input [34]. Thus, a hypothetical tactile contribution to the late evoked responses in S1 cortex, caused by withdrawal reflexes, would not be expected to be blocked by PAG/DRN stimulation. In the present study, by analyzing data recorded under control conditions (i.e. without PAG/DRN stimulation) during weeks 4-5 PS, we confirmed that these responses were still elicited in the instances when cutaneous nociceptive stimulation did not cause a withdrawal reflex (figure S1).

Experimental protocol 2.4.1. Experimental set-up
Timeline of the experiments is summarized in figure 1(B). All sessions lasted maximally 5 h and were performed with a break of at least 24 h in between sessions. For each session, the rats were allowed to get accustomed to the setting for 30 min after connecting the electronics. In each session, the pain threshold on CO 2 laser cutaneous stimulation was defined as the lowest stimulation duration to elicit a withdrawal reflex in the right hind paw in ⩾3/5 tests. The impedances of all implanted electrodes were measured to check their functionality throughout the study (supplementary figure S2). Stimulation electrodes with an impedance below 1 MOhm were considered in electrical contact. The stimulation parameters used were 50 Hz and biphasic charge-balanced squared pulses (2 × 50 µs negative square pulse followed by a positive square pulse). When more than one microelectrode was used for PAG/DRN stimulation, the stimulation pulses were synchronized. Plexon multichannel neural data acquisition system (OmniPlex, Texas, USA, Plexon Inc.,) was used for neural recordings while the neural stimulations were delivered via a current-controlled stimulator (Plexon Inc., Texas, USA, PlexStim v2.2).

Microelectrode selection
During weeks 2-3 PS, we selected the microelectrodes and stimulation parameters in PAG/DRN that blocked the withdrawal reflex without causing noticeable behavioral side effect. This was made by increasing the current in steps of 10 µA (maximum 50 µA, while the stimulation frequency was kept at 50 Hz) for each of the 16 microelectrodes, until behavioral side effects could be detected. Side effects were defined as any visually observable deviation from the regular behavior or normal posture caused by the stimulation (e.g. noticeable increase in breath rate, signs of enhanced vigilance or alertness, aversive responses such as escape and sudden rotations). Different combinations of promising 'analgesic' microelectrodes were then tested to finally select a subgroup and minimal stimulation current that induced complete block of withdrawal reflexes (zero reflexes on 10 CO 2 laser stimulations) without noticeable behavioral side effects. The stimulation current per electrode could often be reduced when using more than one electrode. This selection and stimulation current was then kept in subsequent sessions throughout the study except for when using an adaptive current stimulation protocol.

Stimulation protocols
During weeks 4-5 PS, the animals were placed on a rectangular metal grid (52 × 53 cm) with Plexiglass walls, and we evaluated the analgesic effects on (WR, defined as the number of withdrawals in response to 16 laser CO 2 pulses divided by the total number laser pulses, n = 16). Intracortical responses and WR were assessed during the cutaneous nociceptive stimulations (n = 16), with or without PAG/DRN stimulation using the selected microelectrodes. The WRs and intracortical responses were monitored at regular time points during 4 h PAG/DRN stimulation at constant current stimulation protocol (0, 15, 30, 60, 90, 120, 150, 180, 210, and 240 min from the beginning of the stimulation and 0, 15, and 30 min post-stimulation).
To understand whether an adaptive current stimulation protocol could maintain the analgesic effect over longer periods of time, 4 h of continuous stimulation were repeated in another experimental session with at least 24 h pause from the constant current stimulation protocol. During this test, the PAG/DRN stimulation current was successively increased by 5 µA (ma × 50 µA) from an initial current of median 30 µA per microelectrode whenever >2/16 skin stimulations caused a reflex. This threshold was considered to be a clear indication of loss of analgesic effect. In 30% of the animals, the current per electrode when using >1 microelectrode during the 4 h adaptive current protocol did eventually exceed the threshold for eliciting side effects during single electrode stimulation (in the selection phase). Nevertheless, no side effects were noted during the 4 h stimulations.

Animal tracking and ECoG testings
For behavioral and ECoG documentation, we added a black floor to the cage, and nest materials, water supply or food in three different corners. Video documentation (Logitech HD pro webcam c920, Logitech Inc.) and electrophysiological recordings were made for one hour during control and during PAG/DRN stimulation.

Pre-perfusion stimulation and perfusion
During week 6 PS, we evaluated the analgesic effects of PAG/DRN stimulation on WRs and intracortical responses using 16 laser CO 2 stimulations at pain threshold. The animals were then deeply anesthetized by pentobarbital (150-200 mg kg −1 , i.p.) and transcardially perfused with ∼100 ml saline at room temperature (RT) followed by ∼300 ml of ice-cold 4% paraformaldehyde in 0.1 M phosphate buffer at 7.4 pH. Subsequently, the brains were dissected, put in 20% sucrose solution for cryoprotection and frozen at −78 • C in 2-methylbutane.

Immunohistochemistry
The frozen brains were mounted on sectioning blocks using Richard-Allan Scientific™ Neg-50™ Frozen Section Medium (Thermo Scientific). Sections were cut horizontally using a cryostat (CryoStar NX70, Thermo Scientific) at 16 µm thickness onto Super Frost ® plus slides (Menzel-Gläser, Thermo Scientific). We followed a standard immunohistochemistry protocol as previously described in Mohammed et al 2020 [47]. In short, the sections were rinsed three times in phosphate-buffered saline (PBS). Permeabilization, while preventing unspecific binding, was achieved by incubating the sections in a blocking solution containing Triton X-100 (Sigma-Aldrich, 0.25%) and normal goat serum (Abcam ab7481, 5%) in PBS. The sections were then incubated at RT overnight with primary antibodies diluted in blocking solution.
The following day, sections were rinsed three times in PBS and incubated with secondary antibodies and 4 ′ ,6-diamidino-2-phenylindole (DAPI) in blocking solution for 2 h in the dark at RT. Neurons, reactive astrocytes and activated microglia were visualized with primary and secondary antibodies and nuclei with DAPI (table 1). Sections were subsequently rinsed three times in PBS and coverslipped using polyvinyl alcohol mounting medium with DABCO ® (Sigma-Aldrich 10 981). Histological images were captured (2x and 10x magnification) using a DS-2MV digital camera (Nikon) mounted on a Nikon eclipse 80i microscope or using a DS-2MV digital camera (Nikon) (4x magnification) mounted on a BX53 microscope (Olympus).
For quantification 10x images (2 sections/marker/animal) with the following markers were analyzed: ED1, glial fibrillary acidic protein (GFAP), neuronal nuclei (NeuN) and DAPI. We used the same parameters (gain, contrast, exposure time, etc) for the respective markers and magnifications. All stained sections were screened and two sections per marker were chosen for each animal at the level of PAG, at approximately 300 µm from the distal microelectrode cluster tip.

Data analysis 2.6.1. PAG/DRN histology
The histological quantification method has been described in detail previously [48,49]. In short, for each marker a detection threshold was set at a fixed ratio above the background intensity, to ensure that only specific staining was analyzed.
The detection thresholds were attained empirically by letting two skilled histologists evaluate a representative subset of the stained sections from all analyzed animals. The resulting fixed detection ratios were taken as the average of the two values obtained for each marker. The pixels in each ROI, with an intensity above this fixed ratio, was counted as marker-positive area. The detection threshold was set to 5.7 times the mean background intensity for GFAP immunofluorescence, to 5.9 times for ED1, to 4.4 for NeuN and to 2.9 times for DAPI. The marker-positive area is expressed as a percentage of the area above threshold with respect to the total ROI area. For each marker, two sections per animal at ∼300 µm proximal to microelectrode tips were analyzed (NIS-Elements 3.2 Software, Nikon Instruments). The innermost ROI extended 50 µm from the microelectrode cluster with the area occupied by the 16 microelectrodes (5385 µm 2 ) subtracted. Further analyzed zones were between 50-100, 100-150 and 150-200 µm from the innermost ROI. A rectangular area (44.700 µm 2 ), outside the outermost ROI, was used as a reference (supplementary figure  S3). The probe coordinates were determined from sections (2x magnification) stained with GFAP and DAPI (supplementary figure S4) and visualized in the Waxholm brain atlas [50], using MATLAB (MATLAB, The MathWorks Inc., Natick, MA, USA).

Electrophysiology
Evoked cortical responses were estimated from the evoked FPs and MUA obtained during −0.2-0.8 s relative to onset of the nociceptive stimulation. Individual evoked FP were extracted from the low-pass filtered recordings (<200 Hz, sampling frequency = 1 kHz), pre-processed to suppress artifacts, averaged and smoothed to provide an overall estimate of the evoked FP per channel. Evoked MUA was obtained from detected spike times in the high-pass filtered (>300 Hz) recordings and estimated as the smoothed normalized z score [34]. Spike detection was performed (per channel) by applying a threshold of minus three times the estimated noise level. Spikes with peak-to-peak amplitude >1 mV were assumed to be artifacts and were discarded from the analysis ( figure S5). Due to the presence of spikes (which can be considered to be outliers when estimating the noise level), the noise level was estimated using the scaled median absolute deviation (a robust estimator) in a 5 s-long moving window [51]. In each animal, the mean evoked FP and z score from all channels (on 16 nociceptive stimulations) were used in the following analysis.
An IOI for quantifying the amplitude of late cortical pain responses was automatically determined based on the grand-mean MUA control responses [34]. Within the respective IOI, the amplitude of the FP response was defined as the maximum value of the rectified channel-average response and the MUA amplitude as the average z score of the channelaverage response in each animal.

Motion tracking
The rats' bodies were detected in video frames and tracked over time using a Kalman filter. Occasional tracking errors were corrected manually. The center of the rat's body-position was estimated using a second-degree 1 s Savitzky Golay smoothing filter to determine the speed and zone-presence within the cage. To estimate the percentage of time spent moving, a speed threshold of 1% of the cage diagonal/s was used. Averaged epochs (5 s) per animal were used in the statistical analysis. For each animal, heatmaps of the cage (single pixel resolution) representing the mean animal presence per pixel (frame rate 5/s) were calculated and visualized.

Statistics
Paired statistics was used in which each animal served as its own control, thus reducing the number of animals necessary for adequate statistics. The chosen number of animals, number of observations/animal and number of CO 2 laser stimulations per test were all based on the predicted variations in WR, FP and MUA responses and power of the PAG/DRN stimulation observed in previously published studies [34,40] using similar outcome measures. In each animal and time point, the averaged response from 16 CO 2 laser stimulation was used as the data value. The median of these averaged values was used when pooling and representing data from respective time point from all animals. Evoked responses over a 4 h stimulation session, with the adaptive or constant current protocols, were analyzed by fitting a linear model and a sigmoid function, respectively, to the median (across animals per time point) evoked response amplitudes (FP and MUA) and WR. The linear model was used to test the hypothesis that the response amplitude did not change over time when using the adaptive current protocol. The sigmoid function was used to predict the asymptotic value and convergence time of response amplitudes. Convergence time was taken as the time point when the response amplitude (as modeled by the sigmoid function) had undergone 95% of the total change occurring between 0 and 240 min.
A Wilcoxon matched-pairs signed-rank test was used to compare the response amplitude 30 min post-stimulation with control, and the response amplitudes with or without PAG/DRN stimulation prior to perfusion. The Wilcoxon test was also used for all comparisons of motion tracking parameters between PAG/DRN stimulation and control. The non-parametric Friedman test with Dunn's multiple comparison test was used to compare the immunohistofluorescent markers in four different ROIs with the reference (GraphPad Software Inc., USA). P < 0.05 was considered statistically significant.

Excluded animals and missing data
Compared to the included animals (n = 12), the microelectrodes in one excluded rat were found in a more rostro-dorsal part of PAG, known to provoke aversive reactions [7,55]. ECoG data from one of the included rats were missing due to computer malfunction. In one additional animal, the ECoG, motion tracking data, recordings at week 6 PS and histological data were missing as the animal was prematurely terminated due to accidental probe detachment from the skull. Probe localization was not possible in one additional animal, due to extensive freeze damage. Furthermore, detailed histological quantifications were not possible in six rats due to freeze damage during fixation procedures.

General aspects
Histological assessment of microelectrode placement in PAG/DRN was possible in 10 of 12 rats. In these rats, the stimulation microelectrodes were confirmed to be implanted in the ventral-caudal part of PAG/DRN forming clusters with an outer diameter of ∼600 µm (figure 2(B) and supplementary video (1)). Impedance measurements showed that 189 of 192 implanted stimulation microelectrodes were in electrical contact (median impedance = 126 kOhm, no significant change over time, supplementary figure  S2). The recording microelectrodes implanted in S1 cortex (median impedance = 1.28 MOhm) exhibited no significant change over time (supplementary figure S2).
In each animal, despite the rather wide area of PAG targeted ( figure 2(B)), we were able to select an individualized subset of microelectrodes (n = 1-3, median = 2, usually including electrodes with contacts at two depths, figure 2(C)), that upon stimulation abolished the nociceptive evoked withdrawal reflexes in the right hind paw without causing noticeable behavioral side effects. These subsets of microelectrodes, having a median minimum stimulation current for effective analgesia of 30 µA/microelectrode (range 20-40 µA/microelectrode, stimulation frequency of 50 Hz), were kept throughout the study.

Analgesia during prolonged PAG/DRN stimulation
During week 4-5 PS, the animals were subjected to a 4-hour protocol in an open field in which PAG/DRN stimulation current was either adaptable (incremental increase of 5 µA when >2/16 nociceptive stimulations resulted in withdrawal reflexes) or kept constant at the predetermined minimum stimulation current (as above).
With the adaptive current stimulation protocol, the nociceptive-evoked FP, MUA and WR were almost totally abolished for 4 h (figures 3(A)-(C)). The total current increase needed to sustain complete blockage of nociceptive responses ranged between 5-15 µA ( figure 3(D)). After the PAG/DRN stimulation was interrupted, the analgesic effects declined but was detectable for at least 30 min for the FPs (figures 3(A)-(C)). No behavioral side effects were elicited during the adaptive current stimulation protocol.
With the constant current stimulation protocol, the analgesic effect seemed to start declining around 30 min after the stimulation initiation. For FP ( figure 3(A)), the analgesic effect dropped to about half during the stimulation period (convergence time: 124 min). For MUA and WR (figures 2(B) and (C)), control levels were eventually reached (convergence time: 103 and 179 min, respectively). For the FP, there was a substantial post-stimulation analgesic aftereffect lasting at least 30 min.
An additional short-lasting stimulation session just before perfusion at week 6 PS (one trial only per animal, each trial comprising 16 CO 2 laser stimulations) using the same microelectrode subset and stimulation parameters as defined weeks 2-3 PS confirmed that a potent analgesic effect could still be elicited (figure 4).

Animal motion and ECoG analysis
To further assess possible unnoticed side effects, we analyzed the animals' movements in an open To further assess potential side effects during PAG/DRN stimulation, we also analyzed changes in recorded ECoGs. As can be seen in figure 6, the overall shape of PSDs of ECoGs recorded during the same sessions were very similar in the two conditions, in both cases dominated by low frequency components ( figure 6(A)). Quantitative comparison of the difference in relative band power showed a small, although significant (p < 0.05), difference in the theta (4-10 Hz) and gamma (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45) Hz) bands (−3.1% and +1.4%, respectively, indicating a minor shift in PSD towards higher frequencies; figure 6(B)). This may suggest the induction of a slight shift in the level of wakefulness during stimulation, although not detected in the behavior (figure 5).

Tissue reactions
No signs of bleedings and only small tissue reactions/neuronal loss were seen within and around (extending not more than 100 µm) the stimulation microelectrode clusters in PAG/DRN (n = 10). The detailed quantitative histological analysis (n = 4) showed a statistically significant increase of reactive astrocytes (GFAP) and small, but statistically significant, increase of activated microglia (ED1) in the innermost ROI (0-50 µm) vs reference. Outside this ROI, only GFAP reached significance in the adjacent ROI (50-100 µm). The neuronal density appeared to be smaller in the inner ROI but this difference did not reach statistical significance (likely due to the limited data available) whereas there was no obvious change (D) stimulation current. The data was obtained before, during and after PAG/DRN stimulation (n = 12). The blue and yellow lines represent the data obtained in the two test conditions (constant current and adaptive current stimulation protocol, respectively) as described in the figure. The shaded pink areas represent the test periods without PAG/DRN stimulation (pre-stimulation control and post-stimulation). The dashed lines show the sigmoid or linear function fitted to the constant and adaptive current stimulation protocol, respectively. The red circle indicates the convergence time of the response amplitude for the sigmoid functions (95% of the predicted asymptotic response amplitude). For each condition, the values obtained 30 min after the end of the stimulation were tested against the control ( + P < 0.05; ++ P < 0.01 Wilcoxon matched-pairs signed-rank). The colored lines in the five boxes above the graphs represent the median evoked FPs (A) and z score (B) at the time points indicated above each graph during constant (in blue) or adaptive current stimulation protocol (in yellow), recorded with the microelectrode array in S1. The first dashed vertical grey line shows the nociceptive stimulus onset (t = 0), while the following two lines indicate the selected IOI, corresponding to the late cortical pain response. The shaded area around the median values represents the interquartile range. −A, inverted amplitude; FP, FP voltage; avg. Z-score, averaged score in the IOI; n.s., not significant.
in neural density in the other ROIs (figures 7(A)-(E)). Qualitatively similar results, corroborating the results from the quantitative data analysis, were seen in the six animals excluded from the quantitative histological analyses due to occurrence of freeze damage (figures 7(G) and (H)).

Assessments of pain in animals
Conventional nociceptive reflex tests were used to select an appropriate individual subset of microelectrodes and stimulation parameters. Since reflex tests and pain perception may differ [40,56], we also recorded nociceptive evoked potentials and multiunit responses in S1. This cortical region is known to receive and process nociceptive information from multiple ascending pathways [57], some of which are assumed to be involved in sensory functions. Moreover, nociceptive evoked S1 cortical potentials are known to strongly correlate to perception of pain magnitude in humans [37,39]. By recording analogous and thus translational pain-related signals in awake animals, a higher validity than what can be obtained from nociceptive reflex tests alone, is thus achieved.
Interestingly, while the nociceptive withdrawal reflexes and cortical pain related signals were initially similarly inhibited by the individualized PAG/DRN stimulation, the inhibition of nociceptive reflexes tended to be less robust and showed a more rapid decay of post-stimulation effects, as compared to the inhibition of cortical FPs, during and after the constant current stimulation protocol. Since the nociceptive evoked FPs reflect the spino-thalamic nociceptive input to S1 cortex [58], whereas withdrawal reflexes reflect nociceptive processing in the spinal sensorimotor circuits, this finding may suggest presence of a differential PAG/DRN control of different nociceptive spinal pathways [4,5,59,60]. In view of the present finding that normal behavior is not significantly affected by the granular stimulation of PAG/DRN employed, it may further be speculated that nociceptive defensive motor reactions, similar to other motor programs, can be relatively spared during prolonged PAG/DRN antinociceptive control.

On the concept of individualized granular stimulation
A key feature of individualized granular stimulation is the ability to make use of an individually selected combination of stimulation sites in the target tissue. This selection is enabled by implanting a surplus of microelectrodes. By keeping microelectrodes provoking adverse effects idle during stimulation, an unprecedented positive ratio between therapeutic efficacy and adverse, clinically unacceptable side effects is achieved [34,41]. This strategy is potentially applicable to several brain targets [41]. Moreover, by spreading out the electrode contacts, the problems of not knowing in advance the precise coordinates of therapeutic stimulation sites in an individual, or which sites are to be avoided, are substantially reduced. Importantly, such a cluster of microelectrodes enables a vast number of possible combinations, and therefore possibilities for synergistic effects, of microelectrodes and stimulation parameters. This creates opportunities for finetuning the granular stimulation field in 3D to spatially adapt to the structure of the neuronal networks of the target tissue in the individual.
The present study thus indicates that adverse effects are to a large extent site specific [61] and that even small differences in stimulation sites with respect to the complex anatomy of the neuronal circuits may result in adverse side effects upon stimulation [8,11,34]. Importantly, granular stimulation fields spatially tailored to the architecture of the neuronal networks cannot be provided by electrodes in the millimeter scale currently used for DBS in the clinic [61], which instead creates more uniformly decaying stimulation fields that can be modulated with respect to size and preferential direction from the electrode body by changing the stimulation amplitude [62] and by switching the active contacts on the electrode body [61], respectively. Still, DBS has been very useful clinically, for example in late stages of Parkinson's disease, and the success of DBS when nothing else has worked has highlighted the huge potential of brain stimulation [13,14].

Significance of the absence of noticeable adverse effects during prolonged PAG/DRN stimulation
We found no indications from the ECoG and motion analysis of major alterations in brain state or behavior (such as changed preference for feeding, drinking and nesting area) nor on general motor activity during ongoing (1 h long) stimulation, indicating minimal interference of the individualized granular stimulation in PAG/DRN on normal life, despite the powerful analgesic effects. These findings  significantly add to previous findings of minimal effects on tactile transmission to S1 cortex, minimal effects on spontaneous cortical activity and stable stimulation thresholds over an 11 weeks stimulation period [34]. Hence, by selecting favorable stimulation sites within PAG/DRN, it is possible to avoid side effects while still providing very potent analgesia. The present findings also confirm the presence of highly specific control functions in PAG/DRN, selectively acting on pain related pathways [4,59,60]. The specific effect of individualized granular stimulation contrasts to the results seen after conventional DBS for which side effects is a major challenge [17][18][19] and also to the common occurrence of side effects induced by available centrally acting pharmacological agents, such as sedation and other impairments of cognitive functions [63][64][65].

Significance of minimal tissue reactions
The findings of small tissue responses, absence of bleedings in the target and small loss of nearby neurons at the conclusion of the experiments (week 6 PS, i.e. in the chronic phase after implantation in rats) are encouraging since they suggest that individualized granular stimulation will be able to fulfill the safety requirements for clinical use. Moreover, the higher biocompatibility in comparison to clinically used electrodes, which are known to provoke substantial tissue reactions and neural death [32,33], is likely of importance for achieving effective stimulation at low stimulation currents, which reduces the risk for current spread to sites causing adverse effects. However, it remains to evaluate the tissue responses over longer time periods than studied here to implanted microelectrodes.

On the sustainability of analgesic efficacy of individualized granular stimulation
The present study shows that an adaptive current stimulation protocol employing small incremental increases in steps of 5 µA during the 4 h stimulation period is an effective way to mitigate tendencies for a decay in analgesic efficacy. The mechanisms of the decay are not yet known, but the finding that a small increase in stimulation current sufficed to bring back an almost total inhibition may suggest that a small reduction in neuronal excitability around the microelectrodes is induced by the stimulation [66]. Given that the initial stimulation current was set at a minimum, i.e. just above the threshold for full analgesia, a small decrease in neuronal excitability can be expected to cause a substantial change in analgesic efficacy. Contributing mechanisms may also include fatigue or excitability changes [24,66] in the descending connections to the spinal cord. However, irrespective of precise mechanism, one conceivably effective strategy for future clinical applications using longer periods of stimulation could be based on the present finding that potent analgesia without side effects can be induced from a rather wide area within PAG/DRN ( figure 2(B)). Hence, by designing a probe comprising microelectrodes that are spread in a larger area, it will likely be possible to define several subclusters within PAG/DRN that can provide analgesia without side effects in each individual. This finding thus enables a protocol whereby different subclusters are stimulated in an alternating way, thereby incorporating recovery periods for respective subclusters, while sustaining analgesia. In addition, intermittent complete pauses in the stimulation utilizing the observed post-stimulation after-effects and frequency modulation may be conceived. To further assess the clinical potential of granular stimulation in PAG/DRN, an evaluation of its effects on chronic pain from different body organs will also be needed.

Conclusions
The present study shows that individualized and adaptive current granular stimulation of PAG/DRN enables long-lasting potent analgesia in awake freely moving rats. Moreover, we found minimal influence on behavior and brain state during the individualized granular stimulation. The small glial reactions and small loss of neurons adjacent to the implanted microelectrodes 6 weeks after implantation Indicate that the used microelectrode technology including the adaptive current stimulation paradigm is biocompatible. Hence, these findings taken together indicate that individualized granular stimulation of PAG/DRN has the potential of providing highly potent and safe analgesia in patients. However, further studies, in particular on chronic pain conditions from different body organs, will be needed to assess the clinical potential of granular stimulation in PAG/DRN.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).