All animal experiments were carried out in accordance with the guidelines of the German Animal Welfare Act and approved by the local authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen). Animals were socially housed under a fixed 12:12 h light/darkness cycle with ad libitum access to food and water.

A total of 15 adult, male NMRI-Foxnl^nu/nu mice (12–14 weeks old, 32–37 g; Janvier, France) were used in this study. All surgical experiments were performed under anesthesia with 1.5–2% Isoflurane in a mixture of 70/30% N₂O/O₂ atmosphere. Analgetic treatment included subcutaneous injection of 4 mg/kg Carprofen (Rimadyl, Zoetis, Berlin, Germany) twice a day, for 3 days following surgery. Two days after stroke induction by the filament occlusion method (18, 23), the animals received an intracortical stem cell implantation on the ipsilateral hemisphere. Resting state fMRI was performed 1 week before, as well as 1, 2, 4, 8, and 12 weeks after stroke induction and cell implantation (2 days after stroke induction), while vitality of the transgenic graft was monitored by BLI at weeks 2, 3, 4, 5, 7, 9, 10, and 12 after stroke.

After MRI baseline measurements, focal ischemic stroke was induced by transient occlusion of the middle cerebral artery (MCA) with an intraluminal filament as described previously (23, 25). In brief, mice were anesthetized with Isoflurane and the right common carotid artery (CCA), the external and the internal carotid artery (ICA) were exposed. A rubber-coated filament (length of 20 mm, diameter of 170 μm at the tip; Doccol Corp., Sharon, USA) was inserted into the ICA until blockage of the MCA. After 30 min occlusion time, the filament was removed and the CCA was ligated permanently. Analgesic medication was started directly before surgery and further provided during 3 days thereafter. Location and size of the ischemic damage was assessed at 48 h after stroke induction by MRI.

The commercially available human neural stem cell line H9-NSC (WA09, Life Technologies) was lentivirally transduced with the construct EF1α-Luc2-T2A-eGFP to express the light emitting imaging reporter firefly luciferase (Luc2) and an enhanced green fluorescence protein (eGFP) under the constitutive promotor of the elongation factor 1 alpha (EF1α), as described earlier (26).

Two days after stroke induction and before cell implantation, all animals received a T2-weighted MRI scan to determine successful stroke induction and, based on MRI results, were divided into cortico-striatal and striatal lesion group, respectively. For cell implantation, NMRI-Foxnl^nu/nu mice were anesthetized and the head was fixed in a stereotaxic frame (Stoelting, Dublin, Ireland). A small hole was drilled into the skull above the planned injection site, and a Hamilton syringe (26 G needle) was slowly inserted ipsilesionally into the somatosensory cortex (S1) at the injection site with AP: +0.5 mm, DV: −0.5 mm according to bregma. The lateral coordinate had previously been selected individually based on the respective T2-weighted MRI at 48 h and was set to implant the cells adjacent to but outside the lesion for each individual animal, at approximately L: 1–2 mm.

Nine animals which had a cortico-striatal stroke and six animals with a striatal stroke were injected with a total of 2 μl of H9-EF1α-Luc2-T2A-eGFP cells at a concentration of 150,000 cells/μl, resuspended in Hank's balanced salt solution (HBSS, Life Technologies), at a flow rate of 150 nl/min. After injection, the needle was kept in place for 5 min before withdrawal.

Cell vitality was monitored via BLI, which was recorded once every 1–2 weeks for the duration of the study. For BLI measurements the mice were intraperitoneally injected with 300 mg/kg D-Luciferin sodium salt (Synchem, Felsberg, Germany) solved in Dulbecco's phosphate-buffered saline (Life Technologies) and subsequently anesthetized with a mixture of 2% Isoflurane in 70/30% N₂O/O₂ atmosphere (24). Photon Emission (PE) was recorded for 30 min with the Photon Imager IVIS SPECTRUM CT (Perkin-Elmer, Waltham, MA, USA) under Isoflurane anesthesia.

All experiments were conducted on a small animal 9.4 T horizontal MRI system with a 20 cm bore diameter and actively shielded gradient coils (BGA12S2, >660 mT/m, Bruker BioSpin, Ettlingen, Germany). RF excitation and signal reception were performed with a 1H quadrature cryogenic surface coil (CryoProbe, Bruker BioSpin). Image acquisition was executed with ParaVision 5.1 software (Bruker BioSpin GmbH). Physiological parameters were monitored with the SA Instruments 1025T System (SA Instruments, NY, USA) and recorded with DASYlab Software (Measurement Computing, Norton, USA). Body temperature was measured via a fiber optic rectal probe (SA Instruments, NY, USA) and kept constant at 37°C ± 1.0°C by a water circulating system (medres, Cologne, Germany). Animals were anesthetized with 2% Isoflurane in a mixture of 70/30% N₂/O₂ and the head was fixated in the animal cradle with ear bars and a tooth bar in a nose cone with continuous gas flow. At the beginning of the imaging session, Isoflurane was reduced to 1.5%.

Functional activity was acquired as resting state functional MRI. All MRI sessions were preceded by a three-plane scout scan (Tripilot), adjustment of the RF signal receiver gain, and a FieldMap with consecutive local shim to optimize magnetic field homogeneity and image quality. An anatomical reference scan was acquired with a T2-weighted TurboRARE sequence with a field of view (FOV) of 17.5 × 17.5 mm², 48 contiguous slices of 0.2 mm slice thickness, matrix dimension of 256 × 256, repetition time (TR) = 5,500 ms, echo time (TE) = 32.5 ms, and a RARE factor of 8 with two averages. Moreover, a T2-weighted spin echo sequence was recorded at 48 h after stroke induction, directly before the stem cell grafting, with equivalent parameters.

A bolus of 0.1 mg/kg Medetomidine (Domitor®, Elanco), suspended in 250 μl NaCl, was administered subcutaneously 15–20 min prior to the functional imaging scan, with subsequent reduction of Isoflurane level to 0.5%, following an earlier reported protocol (18, 27). A gradient-echo echo-planar imaging (GE-EPI) sequence was used for rs-fMRI with the following parameters: FOV: 17.5 × 17.5 mm², matrix size: 96 × 96, in-plane resolution: 182 × 182 μm², TR = 2,840 ms, and TE = 18 ms. One hundred and five image sets were acquired with 16 slices each, with slice thickness of 0.5 mm and inter-slice gap of 0.1 mm, recorded non-interleaved and covering the whole forebrain, starting only after a minimal time of 10 min on reduced Isoflurane levels.

All datasets were brain extracted, slice-wise motion corrected with FSL (FMRIB Software Library; http://www.fmrib.ox.ac.uk/fsl) and linearly detrended (28). To allow for comparison of different imaging sessions for analysis of specific Regions of Interest (ROIs), we co-registered our datasets to publicly available mouse brain atlases of the whole brain (29) and the neocortex (30) through the following multi-step process: (1) First, an in-house made MRI mouse brain template was warped to both brain atlases with FSL. (2) The acquired anatomical reference scan from each individual session was linearly co-registered to the in-house mouse brain template and (3) the functional datasets to the anatomical datasets via rigid-body transformation. (4) Finally, the co-registration results themselves were discarded but the transformation matrices from all steps were concatenated, inverted, and then applied to the mouse brain atlas to fit the atlas to the individual original data, thus avoiding deformation bias of the raw data. Six ROIs, or nodes, were extracted from the sensorimotor network, separate for each hemisphere: The primary and secondary motor cortex (M1/M2), the primary somatosensory cortex excluding the limb regions (S1 w/o limbs), the S1 fore- and hindpaw limb regions (S1 limbs), the secondary somatosensory cortex (S2), the thalamus (Th), and the caudate putamen (CPu) (Supplementary Figure 1). Bilateral ROIs were only analyzed until week 2 after stroke induction for accuracy purposes of ROI extraction. For longer periods after stroke, due to brain swelling and/or atrophy, only the contralateral hemisphere was analyzed.

Regression of physiological noise was performed by regressing out the recorded respiratory signals, motion parameters, and drifts up to the second order (28). Furthermore, the data were spatially smoothed in-plane with a Gaussian filter of FWHM = 0.3 mm. To limit contributions from low-frequency signal drifts and high-frequency physiological noise, the time series signal was bandpass-filtered to 0.01–0.08 Hz and normalized. Group-wise full Pearson correlation between pairs of ROIs (i.e., nodes) of the average time series was calculated with FSLNets (v0.6; www.fmrib.ox.ac.uk/fsl) and compiled in matrix form. All correlation values were transformed to z-values by taking the Fisher transformation of the r-values prior to averaging.

Statistical analysis was performed with IBM SPSS 24 software (IBM Corporation, New York, NY, USA), Matlab 2014b (The MathWorks Inc., Natick, MA, USA) and GraphPad Prism v.8.0.2 for Windows (GraphPad Software, La Jolla California USA). Statistical significance levels were set to ^*p < 0.05, ^**p < 0.005.

Forty eight hours after intraluminal filament occlusion of the MCA, all animals received T2-weighted MRI to confirm successful stroke induction. Two types of ischemic lesions were found on the MRI: a small lesion restricted to the striatum (n = 6), and a large lesion encompassing the striatum plus a large part of the cortex (n = 9). Figure 1 presents the ischemic lesion on T2-weighted MRI of a representative mouse of both lesion types.

Forty eight hours after stroke induction, 300,000 neural stem cells, previously transduced to express luciferase under constitutive control, were implanted into the S1 cortex. BLI was performed every 1–2 weeks, starting at week 2 post MCA occlusion. At the time point two weeks after implantation, BLI signal intensity was strong and indistinguishable between both lesion groups. In the group of animals with striatal lesion, BLI signal remained high with little non-significant [F_{(7, 30)} = 4.02] changes throughout the 12 weeks observation. In the cortico-striatal lesion group, BLI signal started to decrease, non-significantly [F_{(7, 59)} = 0.67] between consecutive weeks, shortly after 2 weeks after cell implantation (Figure 2A). The quantitative analysis of the BLI signal intensity of the cell grafts is given in Figure 2B, clearly showing the persistent high level of stable graft viability for the striatal lesion group (black line). The strong signal decrease in the cortico-striatal lesion group to approximately 30% on average of the original average value indicated a vitality loss of a large graft fraction within the first weeks after implantation (red line).

Many investigations have been concerned with functional improvement of the brain after stem cell implantations (4–6, 12, 13, 31). They mostly relied on behavioral evaluations of the functional deficit and improvement, while histological analysis was used for demonstration of existing stem cell grafts and their potential neuronal differentiation (7–11, 32). However, these histological studies are commonly qualitative assessments of the existence of implanted stem cells, without quantification of the fraction of stem cells survived at the end of the observation period and in correspondence to the functional status. This means that those studies missed a clear temporal relationship between the functional improvement and the persistent survival of the graft. This is, however, of high importance as several studies had reported high mortality of engrafted stem cells (33–39). Rather than on qualitative histological assessment, here, we have relied on the well-established quantitative BLI as a tool to quantitatively determine cell viability. As the firefly luciferase which we used as imaging reporter to follow vitality of our engrafted stem cells depends on the presence of ATP and oxygen for the photon generation, the BLI signal intensity is directly proportional to the number of luciferase positive stem cells (40). It should be noted though, that in an earlier report, we had described differentiation of the exact same stem cells, used in the present study, into neurons 3 months after implantation into the corticostriatal lesion model reporting HuNu/NeuN double-positive cells in a qualitative immunohistochemical analysis (18).

We have demonstrated that correct placement of the graft, sufficiently distant to the ischemic lesion, allows long-term and fully quantitative preservation of the graft viability. However, in the case of our cortico-striatal stroke lesion, the stem cells had been implanted too close to the acute lesion rim, resulting in a later substantial loss of viability as the lesion expansion jeopardized the survival of the stem cell majority (18). This was demonstrated by the severe reduction of BLI signal intensity of the stem cell graft. In contrast, in the case of the striatal lesions, the cortical graft preserved its early full level of vitality over the 3 months observation time, as demonstrated with the steady BLI signal over 3 months following engraftment. This finding is in full agreement with earlier reports of the same stem cell line, H9, in NMRI-Foxnl^nu/nu mice. There, a stable quantitative preservation of cortical graft viability in the healthy mouse brain (26, 41) and also in a different cortical stroke model (41) was seen with robust and stable BLI signal intensity for 7 weeks (26) and even for 3 months (41).

Persistent strong reduction of the functional connectivity strength of both hemispheres had been shown by us (18) and others (19–21) for stroke situation without cell treatment. In correspondence with our earlier study (18), we found a transient stabilization of the functional connectivity after cortical stem cell engraftment in animals with large cortico-striatal lesions. The stabilization remained for ~4 weeks when the connectivity strength decreased to approach that of animals with stroke only. Those reductions in functional connectivity affected also the contralateral hemisphere showing the massive far-range effect of the stroke lesion on the neuronal networks. In contrast, our data on the striatal lesions showed a robust stabilization of the functional connectivity by the stem cell graft, remaining stable beyond the 2 month time point, and still being stronger at week 12 than in the case of cortico-striatal lesions. Interestingly, also the selective intermittent inter-hemispheric connectivity decrease, seen in the striatal lesions with stem cells, showed a clear tendency for re-normalization within 2 weeks.

In the large cortico-striatal lesions, the lesion expanded into the adjacent graft location, thereby jeopardizing the graft vitality. When the graft vitality is reduced, the connectivity stabilization is weakened in consequence and is finally lost, here after about 4 weeks. However, some cells were still surviving and had been shown earlier to differentiate into neurons (18). In contrast, in the small striatal lesion, the graft is not threatened by expansion of the ischemic territory into the cortical cell location and vitality of the graft remains stable over the whole observation period of 12 weeks. In consequence, functional connectivity strength remains at the level of control before stroke at least till week 8. Surprisingly, at week 12 the connectivity strength is also substantially weakened in the striatal lesion group, approximating the lower values of the cortico-striatal group—despite persistent vitality. From our findings we conclude that vitality of the stem cell graft is a necessary, but not a sufficient condition for stabilization of functional neuronal networks.

From our present results it becomes plausible that full quantitative graft vitality is required for functional connectivity stabilization. Previously, we had concluded from comparison of the functional connectivity strength of animals after stroke with and without stem cell engraftment that a paracrine effect must be the responsible mechanism for the stem cell mediated functional network stabilization (18). The here observed much delayed functional connectivity weakening at week 12 post-stroke induction and engraftment points to additional effects that further influence the functional networks stability. Late rejection of the graft can be excluded because of persistent high BLI signal intensities. However, interaction of the graft with other mechanisms must be sought that modulate the paracrine effect, which is therapeutically active during the first 2 months. Here, a hot candidate with time dependent activity are the immune cells of which a strong interaction with stem cells has been discussed (42). Microglia and macrophages have been shown to be of anti-inflammatory, protective polarization phenotype in the acute stroke phase and switch to a pro-inflammatory, aggravating polarization phenotype at later time points (43). Although little is known about the immune cells' activity in the chronic phase of 3 months, it may be speculated that the immune cell activity may affect the paracrine effect of the graft by modulating the stem cell mediated secretion of cytokines and growth factors. Then, external modulation of the polarization state of the immune cells, which has previously been demonstrated to have a protective neuronal effect (44, 45) may also contribute to the recovery of the graft's therapeutically active paracrine effect. Future studies may follow this interaction between stem cell graft and immune cells by multicolor BLI (46), allowing simultaneous monitoring of the graft viability and of the immune cell polarization (47), while assessing the effect of polarization status modulation on the functional brain networks over time.

MH was employed part time by Percuros B.V. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.