Reporter gene-engineering of human induced pluripotent stem cells during differentiation renders in vivo traceable hepatocyte-like cells accessible

Highlights • iPSC-derived hepatocyte-like cells (HLCs) rendered traceable in vivo.• Reproducible lentivirus-based gene transfer during the differentiation process.• Protocol and reporter expression did not negatively impact on HLC maturation.• Proof-of-principle shown for whole-body SPECT/CT-afforded HLC in vivo tracking.


Introduction
Orthotopic liver transplantation (OLT) is the core treatment for end-stage liver disease and some metabolic disorders but patient need consistently exceeds donor organ availability.
Primary hepatocyte transplantation (HTx) is a well-established, safe cell therapy benefitting from reduced invasiveness, retention of the native liver architecture and offering repeat use of cryopreserved donor cells (Forbes et al., 2015). Whilst HTx tends to result in relatively shortterm improvement in hepatic function or graft rejection, preconditioning regimes have shown improved engraftment pre-clinically, but require optimisation (Soltys et al., 2017). Success of HTx engraftment has also been strongly linked to cell quality (Donato et al., 2008;Ibars et al., 2016), yet it relies on hepatocytes isolated from livers rejected for OLT (e.g. geriatric donors, prolonged ischaemia, fatty livers). To circumvent HTx quality and supply limitations, human induced pluripotent stem cell (hiPSC)-derived hepatocytes have been considered a potential substitute. Several protocols were developed to differentiate hiPSCs into hepatocyte-like cells (HLCs) (Hannan et al., 2013;Rashid et al., 2010;Si-Tayeb et al., 2010;Song et al., 2009). Moreover, clinical-grade hiPSC lines capable of differentiation into HLCs (Baghbaderani et al., 2015;Blackford et al., 2018;Wang et al., 2015) accompanied by protocols demonstrating clinically relevant scalable HLC or liver organoid production (Takebe et al., 2017;Wang et al., 2017;Yamashita et al., 2018) significantly progressed the clinical realisation of HLC therapies.
Post-transplant evaluations typically rely on blood/serum analyses for soluble factors and liver enzyme activity, offering no information on the in vivo location of transplanted cells and providing only indirect viability information. Alternatively, histology of biopsied tissues demonstrates localised engraftment but is invasive and a risk to both the host and transplanted cells. The option to track engrafted cells would be highly beneficial. Non-invasive wholebody imaging would provide spatiotemporal information about their in vivo location and viability both short and long-term and allow quantitative comparison between different transplantation strategies.
In vivo cell tracking can be achieved by directly labelling cells or by employing reporter gene technology with the latter offering several advantages; (i) the observation period is independent of the contrast agent, i.e. not limited by label efflux or the half-life of a radioisotope; (ii) genetic encoding avoids label dilution phenomena and better reflects cell viability, and also (iii) circumvents complex direct cell labelling procedures and associated toxicities Volpe et al., 2018). Its drawback is the need for genetic engineering. Host reporter proteins are preferable to foreign reporters, which are prone to recognition/destruction by an intact immune system. Importantly, host reporters should not be expressed in the transplanted tissue of interest, and only in a limited number of other host tissues, ideally at low levels to ensure favourable contrast during imaging.
The human sodium iodide symporter (hNIS) is a transmembrane glycoprotein that has been exploited as a radionuclide reporter gene for both single photon computed tomography (SPECT) and positron emission tomography (PET) in a variety of cell tracking settings; including cancer metastasis (Diocou et al., 2017;Fruhwirth et al., 2014;Volpe et al., 2018), migration of mesenchymal stem cells (Dwyer et al., 2011), tracking of hiPSC and cardiac stem cell myocardial infarction models (Templin et al., 2012;Terrovitis et al., 2008), and embryonic stem cell-caused teratomas (Wolfs et al., 2017). hNIS is endogenously expressed at high levels in the thyroid gland and at lower levels in few extrathyroidal tissues (salivary glands, mammary glands, stomach and small intestine) (Portulano et al., 2014). Its function depends on an intact Na + /K + gradient, driven by cellular ATP, and thus it sensitively reports only live cells.
Here, our aim was to develop a protocol for the generation of in vivo traceable HLCs during differentiation, to enable compatibility with the range of transplantation protocols currently utilised in the field and provide a non-invasive approach to optimise HLC engraftment protocols in the future. We assessed the impact of lentiviral gene transfer on HLC maturation and provided proof-of-principle in vivo detection of resultant traceable hNIS-mGFP + HLCs by SPECT/CT imaging.

Reagents and chemicals.
Purchased from Sigma, Thermo-Fisher, Gibco or StemCell Technologies, unless otherwise stated. All cell lines including hiPSC lines have been previously described and were grown as recommended (cf. Supplement). Standard in vitro methodologies including lentivirus (LV) production, flow cytometry, gene expression analysis, secreted albumin and cell viability determinations, cellular radiotracer uptake, and immunofluorescence staining are detailed in the Supplement.

Cell transduction.
Cells were washed with PBS. Viral particles with an estimated MOI of 5* (based on 2x10 6 cells expected per 10cm dish) were diluted in hepatocyte maturation media and added dropwise to cover cells (3mL media/10cm dish). Dishes were left at room temperature for 15 min. 1.5mLfresh medium was added to each dish and cells were incubated overnight in hypoxic conditions (5% CO 2 , 5% O 2 ). 24h later 1.5ml fresh medium was added and after 48h the cells were washed and lifted for either re-seeding/in vitro maturation, in vivo transplantation, or flow cytometric analysis.

Cell preparation for in vivo experiments.
For intraperitoneal injection of pre-labelled cells, cells were first radiolabelled as for in vitro uptake assays with 100kBq 99m TcO 4 -/mL.
Cells were subsequently washed twice with PBS ++ , lifted using TrypLE and resuspended in PBS ++ at 10 7 cells/100µL. Radiolabelled cells were used immediately for injection into animals.
2.5 Animals. NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG; Charles River UK) were maintained under sterile conditions with food and water available ad libitum. All procedures were performed in accordance with all legal, ethical, and institutional requirements. All SPECT/CT data sets were reconstructed using a Monte Carlo-based full 3D iterative algorithm (Tera-Tomo; Mediso). Corrections for attenuation, detector dead time, and radioisotope decay were in place as needed. All images were analysed using VivoQuant software (inviCRO), enabling the co-registration of SPECT and CT images and delineation of regions of interest (ROIs) for quantification of radioactivity. As background can vary in different locations in vivo it is important to consider local/regional thresholding and segmentation. We employed the 3D implementation of Otsu's method (Otsu, 1979) for rendering volumes with radioactivity counts above background. The total activity in the whole animal (excluding the tail) at the time of tracer administration was defined as the injected dose (ID). Radioactivity in each ROI was quantified using VivoQuant software and expressed as standard uptake value (SUV).

Liver histology. Harvested livers were separated into individual lobes, embedded in
optimal cutting temperature (OCT) medium and frozen in isopentane pre-cooled over liquid nitrogen and stored at −80 °C. 10 µm sections were cut with a Cryomatic cryostat (Bright Ltd, Huntingdon, UK), placed on polysine-coated slides and fixed in 4% paraformaldehyde (15min at room temperature (RT)), washed thrice in PBS before blocked for 1h (PBS containing 1.5% bovine serum albumin, 3% donkey serum, 0.1% Triton-X

Generation of in vivo traceable HLCs by transduction during differentiation.
Established protocols to differentiate hiPSCs into HLCs employ a stepwise approach: induction of endoderm, hepatic specification, hepatoblast expansion and hepatic maturation (Hannan et al., 2013;Rashid et al., 2010;Si-Tayeb et al., 2010;Song et al., 2009). We used two hiPSC lines, A1AT z/z RMA B08 (A1AT) and the cGMP-compliant line CGT-RCiB-10 (CGT10), which differentiated into HLCs adopting the expected morphology throughout (Fig.1A). Briefly, tightly packed hiPSC colonies with a high nuclear-to-cytoplasm ratio and prominent nucleoli expanded rapidly from initially seeded colonies following induction of endoderm differentiation. Definitive endoderm was stimulated towards hepatic endoderm and the resulting hepatoblast population rapidly expanded and proliferated creating a confluent monolayer. Subsequently, a polyhedral monolayer of immature HLCs with high cytoplasmto-nuclear ratio was obtained. On day 18, we transduced the differentiating cell lines with LVs containing DNA encoding for the dual-mode radionuclide-fluorescence reporter gene hNIS-mGFP. Transduction at earlier differentiation time points (e.g. day 10) was inefficient due to either very low transduction efficiencies (hiPSC stage) or cell loss during transduction in the hepatoblast expansion stage (high cell turnover; Fig.S1).
We validated that differentiation occurred as expected at two differentiation time points and compared transduced cells to non-transduced populations from the same differentiation batch. The first time point was two days post-transduction (day 20) at the immature HLC stage when HLCs are typically passaged onto collagen-I for further maturation in vitro or lifted for in vivo transplantation. Flow cytometry confirmed hNIS-mGFP expression (Fig.1B) and comparable expression of the hepatic progenitor marker epithelial cell adhesion molecule relative to untransduced control HLCs (EpCAM; Fig.1C). Additional markers of foetal hepatocyte development were quantified by quantitative real-time PCR to assess whether the hepatic lineage was retained: ɑ-fetoprotein (AFP), cytochrome P450 Family protein 3A7 (CYP3A7), and hepatic nuclear factor 4α (HNF4A) (Fig.1D-F). No significant differences between hNIS-mGFP + HLCs and untransduced control HLCs were observed. Similar results were obtained when primary foetal hepatocytes were transduced (Fig.S2). Long-term culture of hNIS-mGFP + A1AT HLCs demonstrated morphology, cell viability, and albumin production to be comparable to untransduced HLCs, as well as reporter expression to be stable for at least 100 days (Fig.S3).
Following passage of HLCs onto collagen-I, we cultured both hNIS-mGFP + HLCs and untransduced control HLCs for two weeks before assessing whether the transduced cell population retained the capacity to mature. Most cells exhibited the typical polyhedral morphology and cells co-expressing albumin and HNF4A, considered the gold standards for bona fide hepatocytes, were seen in abundance along with more immature cells ( Fig.2A). We further observed that the expected membrane expression of hNIS-mGFP was maintained throughout the maturation process with cells capable of co-expressing HNF4A, albumin and hNIS, indicating similar levels of maturation of both populations. To verify this independently, we analysed media concentrations of secreted albumin by ELISA and confirmed them to be similar (Fig.2B). Moreover, expression levels of ALBUMIN and HNF4ɑ mRNA were comparable between hNIS-mGFP + and untransduced HLCs (Fig.2C-D).
Additionally, we analysed gene expression of the asialoglycoprotein receptor isoform 1 (ASGR1; Fig.2E), which is typically found on the cell surface of functional mature HLCs (Peters et al., 2016). All markers we investigated demonstrated that hNIS-mGFP + HLCs and untransduced HLCs did not differ significantly. This provided evidence that our LV transduction strategy did not negatively impact on maturation to the hepatic phenotype in these hiPSC lines. Furthermore, hNIS mRNA expression levels were retained at comparable levels at day 34 relative to day 20 indicating that reporter expression was stable throughout maturation ( Fig 2F).
We next assessed hNIS-mGFP functionality when expressed in transduced HLCs by quantifying cellular uptake of the hNIS radiotracer 99m TcO 4 -. Both mature (differentiation day 34, i.e. two weeks post-transduction) hNIS-mGFP + HLC cell lines readily took up the radiotracer as compared to untransduced cells ( Fig.2G-H). Moreover, 99m TcO 4 uptake was sensitive to competition with the hNIS co-substrate perchlorate, thereby confirming hNIS specificity.

In vivo imaging of hNIS-mGFP + HLCs.
First, we pre-labelled hNIS-mGFP + CGT10 HLCs in vitro with 99m TcO 4 -(16.2±4.1kBq/5x10 6 cells). The cells were harvested, suspended in PBS to a final volume of 50 µL and injected intra-hepatically into NSG mice, which were imaged by SPECT/CT immediately post-injection (Fig.3A). The signal of the pre-labelled hNIS-mGFP + CGT10 HLCs was clearly visible at the site of injection. Image quantification revealed a total amount of 5.2 kBq 99m Tc in the thresholded and segmented location, which had a total rendered volume of 14.3 mm 3 (Fig.3B-C). Notably, we twice used the same amount of hNIS-mGFP + CGT10 HLCs from the same pre-labelled batch, suspended them in a final volume of 50µL saline in 1.5 mL tubes, but placed them as imaging phantoms alongside the animals' rear legs (Fig.3B-C) as quantification controls. The quantified average 99m Tc activity in the phantoms was 6.5 kBq within average rendered volumes of 14.8 mm 3 . The ~18% lower radioactive signal in the segmented image in the mouse as compared to the phantoms indicates cell loss upon in vivo injection, most likely due to cell dispersal.
24h later, the animals were re-imaged by SPECT/CT following intravenous (i/v) 99m TcO 4 administration to verify if intra-hepatically administered hNIS-mGFP + HLCs were detectable in vivo without pre-labelling ( Fig.3A). Due to the short half live of 99m Tc (τ=6.01h), only ~6% of the radioactivity (0.3 kBq) from in vitro labelling was left at the time of i/v administration of 99m TcO 4 -. Control animals received only the i/v administered radiotracer (Fig.3D). 99m TcO 4 uptake by hNIS clearly demonstrated reporter function in vivo in HLCs and, more importantly, that the administered hNIS-mGFP + CTG10 HLCs were detectable in the liver without interference from any organs endogenously expressing mouse NIS (homologous to hNIS; Fig. 3; (Portulano et al., 2014)). In vivo image quantification revealed HLCs uptake of 25.4 kBq 99m Tc in the background-corrected and segmented location, which had a total rendered volume of 4.1 mm 3 , a noticeable decrease in size relative to the previous scan. Imaging data demonstrated that hNIS-mGFP + CTG10 HLCs remained alive for at least 24h post administration, but were undetectable after seven days, most likely due to death in the liver environment.
To verify that this limited cell survival was not due to reporter expression, hNIS-mGFP + and non-reporter expressing control HLCs from the same batch of differentiation were transplanted intra-hepatically on day 23 of the differentiation protocol. We observed cell survival by SPECT/CT imaging for 2 days post transplantation into the liver in the hNIS-mGFP + HLC population (Fig.4A-C). Following animal sacrifice, we detected in excised livers some hNIS-GFP + HLCs (Fig.4D) and confirmed this by histology ( Fig.4E; staining for human albumin in sections from the transplant sites). Notably, both control and traceable HLCs were detected by histology at day 2. Importantly, human albumin (albeit at low levels due to the early time point after transplantation) was found in the serum of both animals transplanted with hNIS-mGFP + and control HLCs (Fig.4F). Radioactivity measurements of the corresponding organs revealed similar levels in organs endogenously expressing NIS and specific uptake in the livers of mice transplanted with hNIS-GFP + HLCs. Similar to the results in Fig.3, neither hNIS-mGFP + nor control HLCs from the same transplant cohorts were detectable at day 6 (neither by in vivo imaging (Fig.4C) nor by any tissue analyses (data not shown)) indicating that both control and hNIS-mGFP + cells exhibit the same limited survival behaviour in vivo in this animal model. Comparatively, similarly engineered liver cancer cells transplanted subcutaneously or intrahepatically and allowed to establish tumours were detectable for at least five weeks ( Fig.S4-S5; endpoints determined by animal regulations and not by detectability), validating our in vivo tracking approach. This suggests that that death of our HLC populations is not due to the method of transplantation, that there is no disadvantage for traceable HLCs, but rather that the long-term tracking capacity of the traceable liver cancer cells is caused by their inherent survival advantage.

Discussion
Multiple studies, both clinically and preclinically, have aimed to improve the efficacy of HTx therapy (Boudechiche et al., 2015;Nagamoto et al., 2016Nagamoto et al., , 2014Soltys et al., 2017;Yamanouchi et al., 2009). Typically, analyses of HTx engraftment are restricted to serial biopsies and/or serum samples without the involvement of non-invasive monitoring. This is especially problematic for research on: the impact of the transplant site on graft retention; the optimal timepoint within differentiation for maximal engraftment/proliferation posttransplant; the therapeutic benefit/success of suspension cell transplants over novel formats (e.g. hepatocyte sheets, liver organoids); or the benefits of repeat infusions/treatment intervals on graft retention/therapeutic effect.
introducing the radionuclide-fluorescence reporter hNIS-mGFP into differentiating immature iPSC-derived HLCs rendering them traceable in vivo. As LVs efficiently infect non-dividing cells, we chose to transduce differentiating cells after the key stages of specification and proliferation when HLCs already exhibited the distinctive polyhedral hepatic morphology.
Transduction during differentiation renders gene transfer compatible with the range of in vivo transplantation protocols in the field, crucially including transplantation of HLCs prior to final maturation allowing completion of maturation post transplantation (cf. Introduction).
Lentiviral transduction results in gene transfer that is not site-specific but also not random. It results in polyclonal populations in respect to the genomic insertion position, which represents a safety concern for genetic engineering of stem cell therapies. Notably, lentiviruses were found to be safer than for example conventional γ-retroviruses (Biffi et al., 2011). Currently, lentiviruses are the tool of choice to produce several clinically approved Another aspect is the differentiation status of the therapeutic cells; the more differentiated and hence less pluripotent, the smaller the overall risk associated with viral gene transfer (cf. we transduced on day 18 when the differentiating cells were already committed to the hepatic lineage and beyond the hepatoblast expansion stage; Fig.1). It is noteworthy that we did not observe any teratoma formation, in contrast we observed HLC loss within less than a week (see below).
Reporter expression was stable and both parts of the fusion reporter hNIS-mGFP were functional (Fig.2, Fig.S3) with hNIS-mGFP + HLCs detectable in vivo post intrahepatic administration (Fig.3). A week post-administration traceable HLCs were undetectable, likely due to HLC death in our animal model, but traceable control tumour cells remained detectable for weeks when administered both subcutaneously and intra-hepatically demonstrating long-term tracking capability and that death of our HLC population is not due to the method of transplantation but rather survival of tumour cells is due to the inherent survival advantage of SK-Hep liver cancer cells (