2.1.

In Vitro Spectral Unmixing

Previously described variations of WT Fluc containing five mutations (F14R, L35Q, V182K, I232K, and F465R)²⁰ that reduced the surface hydrophobicity and conferred significant thermostability were further mutated to red-shift (S284T) mutant x5red (x5r)¹⁸ and to produce a novel blue-shifted (V241I/G246A/F250S) mutant x5green (x5g) bioluminescence spectra. In vitro, the dual-color Fluc pair x5g/r was tested against the bioluminescence spectra of WT Fluc (Promega, Madison, Wisconsin), PpyRE9 Fluc mutant,²² click beetle luciferase CBG99,²³ and Renilla luciferase hRluc²⁴ were acquired with the IVIS Spectrum CT system equipped with 18 emission filter (20-nm bandpass) at 37°C. Each luciferase was expressed by transfection of lentiviral vectors in the human tumor cell line HEK 293T. The bioluminescence signal was corrected for the co-expressed copGFP fluorescence, measured separately with a plate reader. Here, copGFP levels reflect the transfection-based differences in luciferase expression due to a specific viral T2A linker in the bicistronic lentiviral plasmid providing equal expression levels of both transgenes. The normalized in vitro spectra [Fig. 1] showed differences in peak photon emission as well as spectra shape. Notably, x5g produced in our setting the highest total photon flux (area under curve, AUC), compared to CBG99 and Luc2 (1.79- and 3.69-fold, respectively). For x5r, we found the AUC value to be 2.24-fold higher compared to PpyRE9 but only 0.45 times of the total photons from Luc2. The photon flux from cells expressing hRluc cannot be compared directly due to the different substrate (coelenterazine); however, the spectra clearly indicated the strong monophasic blue/green light emission. In contrast, emission spectra of Fluc and click beetle luciferase followed a biphasic spline and in vitro/in vivo emission maxima of 520/560 nm (x5g, CBG99), 560/620 nm (PpyRE9, x5r), and 560/600 nm (Luc2). The spectral separation was higher for x5g/x5r for all measured wavelengths, compared to the other green/red couples CBG99/PpyRE9, x5g/Luc2, and CBG99/Luc2. The second-best separation was achieved by combining x5g and Luc2.

Fig. 1

Comparison of Luc2, CBG99, PpyRE9, x5g, x5r, and hRluc emission spectra in vitro. Plotted is the photon emission from $\mathrm{EF}1\alpha$-luciferase-T2A-copGFP transfected HEK 293T cells recorded with the IVIS Spectrum CT system (mean of $n=4$ to 6). Photon emission has been normalized to the individual copGFP expression. Graph with gray background is a zoom-in for 580 to 700 nm to highlight the differences in the red part of the spectra.

As the next step, we performed spectral unmixing of HEK 293T cells transfected with x5g or x5r Fluc [Fig. 2(a)]. With the automated spectral unmixing algorithm provided by the IVIS software, the green and red emission spectra of x5g and x5r were extracted [Fig. 2(b)] and used for predicting the x5g/x5r ratio in different mixtures [Fig. 2(c)]. The quantitative analysis of the unmixing data revealed high accuracy of predicted x5g ratios in relation to the actual number of plated x5g cells [Fig. 2(d)]. A linear regression analysis showed that for both spectra, the unmixed photon counts were in linear relationship to the amount of Fluc cells for both reporters, x5g and x5r ($R^2=0.991$ $p<0.001$ and $R^2=0.994$, $p<0.001$). For predicting the correct x5g/r ratios, the algorithm resulted in a very low mean absolute deviation (MAD) of $0.02\pm 0.02$.

Fig. 2

Unmixing of x5g and x5r luciferases expressed by HEK 293T cells. (a) Fluorescence microscopy overlay with phase contrast image of cells at 2 days post transfection, used for the unmixing experiment (scale bar $25\mu \mathrm{m}$). (b) Unmixed spectra of x5r and x5g calculated by the unmixing algorithm. (c) Plating scheme for x5g and x5r cells in different numbers ($n=3$) and the corresponding false-colored unmixed results displayed as green/red composite as well as the open filter, x5g, and x5 red unmixed images. (d) Plot of measured unmixed x5g and x5r photon flux plotted against the plated ratio of x5g cells in the mix ($n=3$). Linear fitting for both luciferases underlines high accuracy of unmixed values with plated cells.

The influence of light absorption and scattering on the x5g/r Fluc photon emission was determined with a tissue-like blood agar mix consisting of gelatin, intralipid, and hemoglobin [Fig. 3(a)]. Plated x5g and x5r cells were covered with different amounts of the blood agar mix before imaging in order to mimic cell grafts implanted at different tissue depths. With increasing blood agar volume, the number of detected photons decreased for x5r and x5g with an exponential two-phase decay [Fig. 3(b)]. The photon flux of x5g compared to x5r remained on average 1.79-fold higher, independent of blood agar volume. Thus, under these in vitro conditions, no change in the x5g/r emission ratio was detected and no additional correction factor for the Fluc-specific absorption was calculated.

Fig. 3

In vitro modeling of superficial versus deep x5g and x5r Fluc sources with tissue-like agarose. (a) Different volumes of tissue-like agarose are placed above the HEK 293T cell layer to mimic the overlaying tissue in different thickness. Micrographs represent selected bioluminescence/photo overlays from the x5g and x5r unmixing, respectively. (b) Quantitative analysis and polynomial fit for the decreasing photon emission, detected when more blood agar volume is added.

2.2.

In Vivo Spectral Unmixing

We determined the efficiency and accuracy of x5g and x5r Fluc unmixing in vivo by BLI of transfected HEK 293T cells which were transplanted into nude mouse brains and imaged on the following day. In comparison to other luciferases, x5g and x5r spectra exhibited substantial differences (Fig. 4). The normalized spectra of x5g revealed, in comparison to the green-shifted click beetle luciferase CBG99 at 560 nm, a 2.07-fold higher photon flux. Thus, the green shoulder of the x5g spectrum was much more prominent. The differences between the two red-shifted Flucs PpyRE9 and x5r in terms of total photon flux and shape of the spectrum were much less distinct.

Fig. 4

In vivo emission spectra of x5g and x5r, in comparison to Luc2, CBG99, PpyRE9, and hRluc, expressed in mammalian cells implanted in the mouse brain (data in part is adapted from our previous publication comparing different luciferases²⁵).

To identify the influence of transplantation depth, we performed experiments with two separate groups: the cortical ($-1.5\text{-}\mathrm{mm}$ depth) as well as striatal ($-3.0\text{-}\mathrm{mm}$ depth) grafts. The experimental setting [Figs. 5(a) and 5(d)] included mice with x5r and x5g cells only and mice that received both cell types in different mixtures. The two control mice were used to extract the in vivo x5r and x5g spectra from the cortex and striatum groups in the auto-unmixing mode, in which the software automatically detected x5r and x5g [Figs. 5(b) and 5(e)]. Notably, the total photon flux of x5g and x5r luciferase in the cortex was 1.93-fold and 1.76-fold higher, compared to the striatum. The emission maximum of x5r in cortical and striatal grafts was 620 nm, however reached in the striatum only 58.44% photon flux. In the case of x5g, the spectrum in the striatal group was 20 nm shifted and the maximal photon emission was only 41.98% of the cortex group. In the cortex group, automatic unmixing of the x5g ratio worked well with high accuracy for all the expected x5g ratios [$F(\mathrm{4,13})=217.455$, $p<0.001$]. A general comparison of expected to calculated x5g ratio resulted in a significant effect [$F(\mathrm{4,17})=79.961$, $p<0.001$], however, with stronger linear correlation compared to the striatum ($R^2=0.994/p<0.001$ versus $R^2=0.933/p=0.008$) [Figs. 5(c) and 5(f)]. A post-hoc comparison showed that, for example, the unmixing result was not precise enough to significantly separate cell grafts with 50% and 75% x5g cells, respectively ($p=0.179$). To further calculate the accuracy of unmixing, the linear regressions for cortical and striatal grafts were statistically compared to the implanted x5g ratios [Figs. 4(c) and 4(f)]. The difference between the slopes for cortical and striatal grafts were not significantly different ($p=0.319$ and $p=0.450$). In addition, a one-sample $t$-test was calculated to compare the unmixing results with the known x5g ratio and MAD was calculated to forecast the error in an experiment. For cortical and striatal grafts, all unmixing ratios were not statistically significantly different except 100% x5g, which might be influenced by the small variation and sample size. The MAD for cortical grafts was smaller (4.5) compared to striatal grafts (8.0), supporting the observation that unmixing results and the expected x5g ratios were found to be very reliable for the cortex and less reliable for the striatum (Table 1).

Fig. 5

Spectral unmixing efficiently determines x5g/r cell ratios in vivo. Transiently transfected HEK 293T cells expressing x5g or x5r luciferases are transplanted into the cortex (a) or striatum (d) and BLI is applied 1 day later. Non-normalized spectra from the x5g and x5r cells differ, dependent on transplantation depth: cortex (b), striatum (e). Spectral unmixing of x5g ratio is in good agreement with expected (dotted line) x5g ratio for the cortical (c) as well as striatal (f) grafts. Significant differences are highlighted by asterisks.

Table 1

Summary of in vivo unmixing validation.

We further processed the data assuming a scenario where the two control mice with x5g and x5r cells for the auto-unmixing mode were not present. The in vivo spectra and mean attenuation factors calculated from three (cortex) or four (striatum) independent experiments were used to generate “library” spectra and to guide the unmixing algorithm [Figs. 5(b) and 5(e)]. In this case, the two control mice x5r and x5g would become obsolete. However, our calculations revealed less precise spectral unmixing for cortical and striatal grafts (Table 1). In the cortex group, statistically significant differences in x5g ratios were not predicted for 25% versus 50% versus 75% x5g cells in the mix. Most obvious was the strong under-representation of x5g in the mix especially for the striatal grafts which led to false unmixing results, e.g., for 50% x5g in the striatum $23.0\%\pm 4.1\%$, which was statistically significantly different from the implanted 50% x5g cells. Similarly, the linear relationship for the increasing ratio of x5g cells in the mix held true but with inferior correlation compared to the automated unmixing (cortex: ${\mathrm{R}}^2=0.939/p=0.007$, striatum: $R^2=0.856/p=0.024$).

2.3.

Discriminating Two Different Neural Stem Cell Populations

To further probe the spectral unmixing approach, we tested x5g in combination with the widely used Fluc Luc2 in NSCs. The human H9-NSC line was stably transduced using $\mathrm{EF}1\alpha \text{-}\mathrm{Luc}2\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{copGFP}$ or DCX-x5g-T2A-EGFP lentiviral vectors and sorted by FACS for their GFP expression, respectively (data not shown). The resulting two stable transgenic cell lines were transplanted into the cortex of nude mice in the same mixture ratios as for the HEK 293T experiment (0/100, 25/75, 50/50, and 100/0% x5g/r) and were imaged 1 and 2 days later [Fig. 6(a)]. The quantitative analysis revealed a linear correlation for both time points with a stronger correlation on the first day after transplantation [Fig. 6(b)]. The presence of the different cell types was verified by immunohistochemistry [Fig. 6(c)].

Fig. 6

Spectral unmixing of two NSC populations. (a) NSC lines expressing x5g or Luc2 are implanted in different ratios into the cortex of nude mice. (b) Quantitative analysis of spectral unmixing from imaging data acquired 1 day and 2 days post implantation (dpi). (c) Representation of immunohistochemical staining of a NSC graft ($\mathrm{EF}1\alpha \text{-}\mathrm{Luc}2\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{copGFP}$). Human nuclei (HuNu) to discriminate the human cell graft in the mouse brain. Enhanced copGFP with anti-copGFP antibody to visualize transplanted cells expressing the transgene. Hoechst stain helps to visualize cell nuclei. Scale bar 50 um.

5.1.

Mutagenesis

Thermostable x5 Fluc contains five mutations (F14R, L35Q, V182K, I232K, and F465R) that reduce surface hydrophobicity and confer significant thermostability.²⁰ The sequence of human codon-optimized WT Fluc is constructed by gene synthesis by oligo assembly and the five mutations are added by splicing by overlap extension PCR-based cloning to produce the plasmid MP5556x5, cloned into SFG-retrovirus³⁵ upstream an internal ribosome entry site (IRES) and a truncated CD34 (dCD34) marker gene. Into this, mutations are added to produce either blue-shifted (V241I/G246A/F250S) or red-shifted (S284T) bioluminescence spectra in mutants x5green (x5g) and x5red (x5r), respectively.

5.2.

Cell Culture

The human embryonic kidney cell line 293T (HEK 293T) was cultured in DMEM + GlutaMAX medium (Life Technologies, Darmstadt, Germany) supplemented with 10% FBS (Life Technologies) and 1% penicillin/streptomycin (Life Technologies) under humidified conditions at 37°C and 5% ${\mathrm{CO}}_2$. Cells were passaged every 3 days and detached with trypsin (Life Technologies). The human NSC line H9-NSC (Life Technologies), which was initially derived from the human embryonic stem cell line H9—NIH Registry WA09,³⁶ was maintained according to the manufacturer’s protocol. Briefly, cells were plated on Geltrex coating at a density of $5\times {10}^4\text{cells}/{\mathrm{cm}}^2$ in StemPro NSC SFM complete medium consisting of $1\times$ KnockOut DMEM/F-12, 2 mM GlutaMax, $20\mathrm{ng}/\mathrm{mL}$ bFGF and EGF, and 2% StemPro supplement (Life Technologies). Cells were passaged every 3 days and detached with StemPro Accutase (Life Technologies).

5.3.

Tissue-like Phantoms

Tissue-like phantom was prepared as follows: $8.76\mathrm{mg}/\mathrm{mL}$ hemoglobin (Sigma-Aldrich) and $40\mathrm{mg}/\mathrm{mL}$ blood agar base (Sigma-Aldrich) were dissolved in 50 mM TBS buffer on a hotplate stirrer at 50°C. When both components were dissolved, Lipovenös MCT (Fresenius Kabi, Bad Homburg, Germany) was added at a final concentration of 1%. For the construction of the tissue phantom, the 96-well plate with the attached transfected cells was placed on a 37°C warming pad, the medium was removed, and the agar mixture was layered carefully above the cells in various volumes ranging from 0 to $175\mu \mathrm{L}$.

5.4.

Cell Transfection and Transduction

HEK 293T cells were transfected with the plasmids $\mathrm{pcDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{x}5\mathrm{g}\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{tdtomato}+\mathrm{SV}40\mathrm{Zeo}$ or $\mathrm{pcDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{x}5\mathrm{r}\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{copGFP}+\mathrm{SV}40\mathrm{Zeo}$, which consisted of the constitutive promoter elongation factor 1 alpha ($\mathrm{EF}1\alpha$), the genetically engineered x5 Fluc red ($+\mathrm{S}284\mathrm{T}$)^18,19,29 and the novel green ($+\mathrm{V}241\mathrm{I}/\mathrm{G}246\mathrm{A}/\mathrm{F}250\mathrm{S}$) mutant, the self-cleaving 2A-like peptide sequence from Thosea asigna virus (T2A), the green fluorescent protein copGFP from Pontellina plumata (Exc 482 nm, Em 502 nm) or the red fluorescent protein tdTomato (Exc 554 nm, Em 581 nm), the Simian virus 40 (SV40) origin, and a Zeocin resistance gene (Zeo). Molecular cloning of plasmids carrying Luc2, CBG99, PpyRE9, and hRluc was described previously.²⁵ For this study, we designed novel bicistronic plasmids for constitutive and cell-specific expression of x5g and x5r by amplifying recombinant DNA by PCR using specific primers bearing appropriate restriction sites in the following steps. At first the backbone $\mathrm{pcDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{MCS}\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{copGFP}$ (System Biosciences, Mountain View) was changed to $\mathrm{pCDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{MCS}\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{tdTomato}$ (pQC NLS TdTomato IX was a gift from Connie Cepko; Addgene plasmid #37347). In addition, we added the antibiotic resistance gene zeocin controlled by the independent SV40 promoter amplified from pBabe Zeo (pBABE zeo Ecotropic Receptor was a gift from William Hahn, Addgene plasmid #10687). Second, x5g or x5r were cloned with BamHI/NotI from M5556X5 or M5549X5 into $\mathrm{pcDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{MCS}\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{tdtomato}+\mathrm{SV}40\mathrm{Zeo}$ or $\mathrm{pcDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{MCS}\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{copGFP}+\mathrm{SV}40\mathrm{Zeo}$. For the cell-specific plasmids, fluorescence proteins were exchanged to tdTomato and EGFP (pmEGFP-1 was a gift from Benjamin Glick, Addgene plasmid #36409). The $\mathrm{EF}1\alpha$ promoter was exchanged by the human DCX promoter (kind gift of Sebastien Couillard-Despres, Paracelsus Medical University, Salzburg, Austria) via ligation with the ClaI and XbaI restriction sites. The final plasmids used for transduction of H9-NSCs were $\mathrm{pcDH}\text{-}\mathrm{EF}1\alpha \text{-}\mathrm{Luc}2\text{-}\mathrm{T}2\mathrm{A}\text{-}\mathrm{copGFP}$ and pcDH-DCX-x5green-T2A-EGFP. Successful cloning was verified by restriction analysis and sequencing.

For transient transfection, HEK 293T cells were seeded ($\mathrm{71,400}{\text{cells}/\mathrm{cm}}^2$) on gelatin-coated P60 cell culture dishes 12 h before transfection. Cells were transfected with $5.25\mu \mathrm{g}$ DNA and $10.5\mu \mathrm{L}$ Metafectene (Biontex, Munich, Germany) in 2 mL Opti-MEM medium (Life Technologies) and incubated for 4 h under normal conditions. Post transfection, cells were cultured for 2 to 3 days and prepared for in vitro dilution series or in vivo implantation.

Lentiviral vector-mediated transduction of H9-NSCs was performed using third-generation helper plasmids and subsequent cell incubation with unconcentrated pseudoviral particles for 24 h (for experimental details, see Ref. 12). We selected stably expressing cell lines based on the LTR-mediated background expression of EGFP and copGFP in the transduced cells via FACS (FACSAria III, BD Biosciences, San Jose).

5.5.

Cell Transplantation

All experiments were conducted according to the guidelines laid out in the German Animal Welfare Act and were approved by the local authorities. Animals were kept in 12 h/12 h day/night cycle in individually ventilated cages. Food and water were offered ad libitum.

Cells transplantation was performed as described before.⁹ Transfected HEK 293T cells were implanted individually or mixed ($\mathrm{x}5\mathrm{r}:\mathrm{x}5\mathrm{g}$; $1:1$, $1:3$ and $4:1$), diluted with a final concentration of $\mathrm{150,000}\text{cells}/\mu \mathrm{L}$ in HBSS (Life Technologies) and stored on ice until implantation. Cells were implanted into the right striatum (AP $+0.5$; L $+2.0$; DV $-3.0$ relative to bregma) or cortex (AP $+0.5$; L $+2.0$; DV $-1.5$ relative to bregma) of NMRI-Foxn1nu/Foxn1nu mice (age 9 to 11 weeks, 25 to 30 g, male from Janvier, Saint Berthevin Cedex, France). In total, we used 28 mice for cortical and 26 for striatal HEK 293T grafts of which 6 and 3, respectively, were excluded because of visual bleeding out of the bore hole.

5.6.

In Vitro and In Vivo Imaging Setup

BLI and fluorescence imaging experiments were performed with the IVIS Spectrum CT (PerkinElmer, Waltham). We recorded emission spectra explicitly with the IVIS system as we used the system throughout all experiments and it was shown by Zhao and colleagues previously that recording emission spectra with the IVIS System resulted in comparable spectra acquired with a spectrophotometer.²³

For spectral unmixing of the in vitro dilution series, respective cell numbers were diluted in PBS (Life Technologies) and plated on a black clear-bottom 96-well plate (Sigma-Aldrich, Taufkirchen, Germany). D-luciferin potassium salt (Synchem, Felsberg, Germany) was dissolved in PBS and added in a final concentration of 1 mM ($\sim 0.3\mathrm{mg}/\mathrm{mL}$). Immediately after D-luciferin application, in vitro dilution series was imaged with following settings: FOV 13.2 cm, excitation block, emission 500 to 840 nm (20-nm bandpass filter) and no filter (open-filter setting), exposure time automatic, binning 8, and f/stop 1. In addition, copGFP fluorescence was measured using the plate reader Mithras² LB943 (Berthold, Bad Wildbach, Germany) with the following settings: excitation 469 nm, emission 510 nm, lamp energy 40%, and counting time 2.0 s. In vivo BLI was performed 1 day post implantation. We used our mouse brain-adapted imaging protocol, which is described in detail elsewhere.⁹ About $50\mathrm{mg}/\mathrm{mL}$ D-luciferin sodium salt (Synchem) was dissolved in PBS, sterile-filtered, and injected intraperitoneally to each animal with a final dose of $300\mathrm{mg}/\mathrm{kg}$ body weight before isoflurane (ISO) anesthesia (2% ISO in 30% ${\mathrm{O}}_2/70\%$ ${\mathrm{N}}_2\mathrm{O}$ atmosphere). Image acquisition was started 8 min post D-luciferin injection for 25 min under the following setting: FOV 22.5 cm, excitation block, emission 500 to 840 nm with 20-nm bandwidth, and open filter. The exposure time ranging from 1 to 60 s, the lens aperture size, and the binning were set automatically by the system to reach the most sensitive setting. Finally, animals were sacrificed and brain tissues were processed for histology.

5.7.

Immunostainings

For immunohistological staining of brain tissues, mice were transcardially perfused under ISO anesthesia with 20 mL PBS followed by 20 mL 4% phosphate-buffered paraformaldehyde (PFA). Brains were removed and frozen in $-40°\mathrm{C}$ cold 2-methylbutane and stored at $-80°\mathrm{C}$. Brains were cut in $14\text{-}\mu \mathrm{m}$ thick sections in coronal plane with the cryostat (Leica, Wetzlar, Germany), mounted on object slides, and stored at $-20°\mathrm{C}$. To equilibrate tissue, brain sections were washed three times with PBS. Antigen retrieval was performed by incubating the brain sections in 10-mM sodium citrate buffer (pH 6.0) at 80°C for 30 min. Unspecific antibody binding was reduced by blocking with 5% normal donkey serum and 0.25% Triton X-100 in PBS for 1 h. Primary antibodies were diluted with 0.25% Triton X-100 in PBS and incubated overnight at 4°C. Brain sections were stained for mouse anti-HuNu (Chemicon, $1:200$) and rabbit anti-copGFP (Evrogen, $1:200$). Secondary antibodies—donkey antimouse-Cy5 (Jackson ImmunoResearch, $1:200$) and donkey antirabbit-Cy3 (Jackson ImmunoResearch, $1:200$)—were diluted with 0.25% Triton X-100 and Hoechst 33342 (Sigma–Aldrich, 1:000) in PBS and incubated for 2 h. Finally, brain sections were air-dried and mounted within Cytoseal XYL (Thermo Fisher Scientific).

All images were acquired with the BZ-9200 Microscope (Keyence, Neu-Isenburg, Germany). Representative images of the immunohistological staining were acquired with the $4\times$ and $60\times$ objectives and the respective phase contrast and fluorescence filter sets (DAPI-BP, GFP-BP, TexRed, and Cy5 HC; Keyence).

5.8.

Data Analysis

Optical imaging data analysis and spectral unmixing was performed with the Living Image 4.3.1 software (PerkinElmer). Regions of interest with constant size were manually drawn for well plate quantification as well as for in vivo data. Total photon flux was calculated as AUC, as described previously,⁹ and the filter with highest average radiance was selected as emission with maximum photon flux. The normalized and non-normalized master spectra for x5g and x5r in the striatum or cortex were calculated by the average of four or three different acquisitions. The master spectrum was used in the “library” mode for spectral unmixing as guideline. The automatic (“auto”) spectral algorithm was used whenever possible, or replaced by the (manual) guided unmixing in which the user has to predefine where to find the pure x5g and x5r spectra. The data for the unmixed populations of x5g and x5r were corrected for the experiment-specific filter attenuation factor. This factor compensated for the difference between the average radiance of x5g/r unmixing at a certain filter (usual 600 to 620/620 to 640 nm) and the open-filter (without filter) setting. Furthermore, this correction factor graded the experimental variability induced by the differences during transfection and transplantation. Without correcting for filter attenuation, the unmixing results will remain qualitative. The filter attenuation factor average of all processed cortical and striatal grafts was used for the master spectrum unmixing.

Histological images were processed with ImageJ 1.48s (National Institutes of Health, Bethesda). Further calculations, plotting, and statistical analysis were done with MS-Excel 2010 (Microsoft Corporation, Redmond) and SPSS 22.0 (IBM, Ehningen, Germany). For multiple group comparisons the one-way ANOVA with Welch statistics and post-hoc pairwise comparison (Tukey-corrected) was used. A $p$-value of $\le 0.05$ was considered to be significant. A Student’s $t$-test was used for the comparison of experimental groups. The one-sample $t$-test was used to compare unmixing results with the implanted x5g ratio. The MAD was the difference between the actual values and the known values (implanted x5g cells). All values were expressed as $\mathrm{mean}\pm \mathrm{standard}\mathrm{deviation}$ (SD) and statistical significance were indicated by *$p<0.05$, **$p<0.01$, ***$p<0.001$, and ^#n.s.