Humanised Mice and Immunodeficient Mice (NSG) Are Equally Sensitive for Prediction of Stem Cell Malignancy in the Teratoma Assay

The use of human pluripotent stem cells (hPSCs) in regenerative medicine has great potential. However, it is important to exclude that these cells can undergo malignant transformation, which could lead to the development of malignant tumours. This property of hPSCs is currently being tested using the teratoma assay, through which cells are injected into immunodeficient mice. Transplantation of stem cells in immunocompromised recipient animals certainly has a much higher incidence of tumour formation. On the other hand, the results obtained in immunodeficient mice could indicate a risk of tumour formation that is practically not present in the human immunocompetent recipient. The presence of a humanised immune system might be more representative of the human situation; therefore, we investigated if the demonstrated malignant features of chosen and well-characterised stem cell lines could be retrieved and if new features could arise in a humanised mouse model. Hu-CD34NSGTM (HIS) mice were compared side by side with immunocompromised mice (NSG) after injection of a set of benign (LU07) and malignant (LU07+dox and 2102Ep) cell lines. Analysis of the tumour development, histological composition, pathology evaluation, and malignancy-associated miRNA expression levels, both in tumour and plasma samples, revealed no differences among mouse groups. This indicates that the HIS mouse model is comparable to, but not more sensitive than, the NSG immunodeficient model for studying the malignancy of stem cells. Since in vivo teratoma assay is cumbersome, in vitro methods for the detection of malignancy are urgently needed.


Introduction
Human pluripotent stem cells (hPSCs) are defined by their capacity to self-renew and to differentiate towards derivatives of the three germ layers: ectoderm, mesoderm, and endoderm. As such, hPSCs offer great promise in personalised regenerative medicine, according to which patients' somatic cells can be reprogrammed in vitro and used to study specific drug responses or disease mechanisms. Furthermore, after correction and differentiation, they could potentially be used in autologous transplantation to improve organ efficiency or substitute damaged tissues [1]. However, before prospective use in the clinic, all hPSCs need to be examined for potential malignant potential.
While pluripotency and developmental potential can be tested in silico with transcriptome analysis, by using tools such as PluriTest [2] or in vitro in short term differentiation experiments (in monolayer culture or embryoid bodies), combined with bioinformatic analysis such as Scorecard [3,4], malignancy can be assessed in vivo only by injection of

The Time of Tumour Development Is Similar between Animal Groups
We compared tumour formation from known benign hiPSCs (LU07), malignant hiPCSs (LU07+dox), and embryonal carcinoma cells (2102Ep) in the teratoma assay, following a well-established protocol [18]. Each cell line was injected into the flanks of three mice per cell line to generate tumours (Figure 1). The teratoma assay is commonly performed with male NSG mice, while HIS mice are delivered as females. Therefore, we compared the tumour formation in three groups of animals: NSG males, NSG females, and HIS females (total n = 27) ( Figure S1). We monitored all animals for tumour growth and collected plasma samples for miRNA analysis (Figure 1). Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW 3 of 17 associated with malignancy, do not differ between NSG immunodeficient mice (NSG) and HIS mice with a reconstituted human immune system.

The time of Tumour Development Is Similar between Animal Groups
We compared tumour formation from known benign hiPSCs (LU07), malignant hiPCSs (LU07+dox), and embryonal carcinoma cells (2102Ep) in the teratoma assay, following a well-established protocol [18]. Each cell line was injected into the flanks of three mice per cell line to generate tumours (Figure 1). The teratoma assay is commonly performed with male NSG mice, while HIS mice are delivered as females. Therefore, we compared the tumour formation in three groups of animals: NSG males, NSG females, and HIS females (total n = 27) ( Figure S1). We monitored all animals for tumour growth and collected plasma samples for miRNA analysis (Figure 1). Figure 1. Experimental setup. Schematic overview of the experimental timeline. Each mouse (NSG and HIS) (n = 27) was injected with 10 6 cells in Matrigel and culture medium (T0) and was weekly monitored for tumour growth. Tumour size was measured using a digital calliper. When the tumour reached ~2 cm 3, it was surgically removed (T1), and the animal was monitored for possible further tumour development till the end of the experiment (T2) when the animals were sacrificed. Plasma samples were collected before cell injection, one week after injection, and subsequently bi-weekly.
After cell injection, the mice were analysed for tumour growth; when a tumour reached ~2 cm 3 , it was surgically removed (T1). Tumours derived from 2102Ep cells grew on average for 39 days in NSG males and 40 and 35 days in NSG females and in HIS females, respectively, before surgical removal (T1) (Figure 2A). In the case of LU07-derived tumours, the times of growth before the removal were 56, 50, and 53 days after cell injection for NSG males and females and HIS females, respectively ( Figure 2B). Tumours generated from LU07+dox cells ( Figure 2B) were removed after 59, 56, and 51 days on average in NSG males, NSG females, and HIS females, respectively ( Figure 2C). There were no significant differences in tumour growth (time from cell injection to removal of the tumour, T1) between all animal groups ( Figure 2).
We also assessed the time the tumours were first detectable after cell injection. The tumours that arose from 2102Ep ECs were first detected on day 22 in NSG males, and on day 21 in NSG and HIS females, after injection of stem cells ( Figure S2). Tumours generated from LU07 hiPSCs were on average first detected on day 34 in NSG males, and days 27 and 34 in NSG and HIS females, respectively, while tumours from LU07+dox hiPSCs were first palpable by days 41, 27, and 34 in NSG males, NSG, and HIS females, respectively. There were no significant differences in the interval between injection of cells and the first detection of the tumour among NSG male, NSG female, and HIS female mice, per cell line ( Figure S2) Figure 1. Experimental setup. Schematic overview of the experimental timeline. Each mouse (NSG and HIS) (n = 27) was injected with 10 6 cells in Matrigel and culture medium (T0) and was weekly monitored for tumour growth. Tumour size was measured using a digital calliper. When the tumour reached~2 cm 3 , it was surgically removed (T1), and the animal was monitored for possible further tumour development till the end of the experiment (T2) when the animals were sacrificed. Plasma samples were collected before cell injection, one week after injection, and subsequently bi-weekly.
After cell injection, the mice were analysed for tumour growth; when a tumour reached~2 cm 3 , it was surgically removed (T1). Tumours derived from 2102Ep cells grew on average for 39 days in NSG males and 40 and 35 days in NSG females and in HIS females, respectively, before surgical removal (T1) (Figure 2A). In the case of LU07-derived tumours, the times of growth before the removal were 56, 50, and 53 days after cell injection for NSG males and females and HIS females, respectively ( Figure 2B). Tumours generated from LU07+dox cells ( Figure 2B) were removed after 59, 56, and 51 days on average in NSG males, NSG females, and HIS females, respectively ( Figure 2C). There were no significant differences in tumour growth (time from cell injection to removal of the tumour, T1) between all animal groups ( Figure 2).  For all animals, the tumours were allowed to develop until they reached ~2 cm 3 before they were surgically removed (T1, Figure 1). The experiments were continued for several weeks to monitor if any of the animals developed tumours after T1 was removed. There were no significant differences between NSG males, NSG females, and HIS females.
We also assessed the time the tumours were first detectable after cell injection. The tumours that arose from 2102Ep ECs were first detected on day 22 in NSG males, and on day 21 in NSG and HIS females, after injection of stem cells ( Figure S2). Tumours generated from LU07 hiPSCs were on average first detected on day 34 in NSG males, and days 27 and 34 in NSG and HIS females, respectively, while tumours from LU07+dox hiPSCs were first palpable by days 41, 27, and 34 in NSG males, NSG, and HIS females, respectively. There were no significant differences in the interval between injection of cells and the first detection of the tumour among NSG male, NSG female, and HIS female mice, per cell line ( Figure S2).

LU07 hiPS Form Benign Teratomas, LU07+dox, and 2102Ep Form Malignant Tumours
For all animals, the tumours were allowed to develop until they reached~2 cm 3 before they were surgically removed (T1, Figure 1). The experiments were continued for several weeks to monitor if any of the animals developed tumours after T1 was removed.
Mice injected with the 2102Ep cell line (n = 9/9) developed T1 tumours and smaller tumours at the end of the experiment (T2; 8/9) ( Figure S3A). Histological analysis revealed typical morphology that is characteristic of EC in both T1 and T2 tumours. All tumours were composed of cells expressing OCT4 and CD30, confirming that these were indeed EC ( Figure S3B,C).
All NSG males, NSG, and HIS females injected with LU07 hiPSCs formed T1 tumours. These were teratomas with derivatives from all the three germ layers, where ectoderm was often represented by neural tubes, mesoderm by various stages of cartilage development, and endoderm by different types of epithelium ( Figure 3A). Unlike 2102Ep cells, mice injected with LU07 did not give rise to new tumours after T1 removal ( Figure S3A).
LU07+dox hiPSCs also generated tumours containing derivatives of the three germ layers ( Figure 3B) but additionally contained areas with EC components characterised by the expression of OCT4 and CD30 ( Figure 3D). The presence of EC components was consistent in all NSG males, NSG, and HIS females. We examined sections of the lymph nodes (subiliac and mesenteric, liver, lungs) by H&E staining and found no signs of metastasis in any of the organs (data not shown).
Teratomas tend to be cystic tumours, in which cysts containing fluid are in between solid areas ( Figure S4). Possibly, the cystic structures hamper the determination of the actual tumour mass. Therefore, the solid surface areas in tumours generated by LU07 and LU07+dox in NSG male and females and in HIS female mice were determined in serial sections (1:10) from all available slides stained with H&E. The total volume of the solid components of the LU07 and LU07+dox tumours (excluding empty spaces generated by cysts) did not differ significantly between NSG and HIS females ( Figure S4).

MiR371a-3p and miR373-3p Indicate the Presence of the Malignant EC Components in Tumours of Both NSG and HIS Mice
We previously [13] showed that after the injection of malignant hiPSCs in mice, high levels of miR-371a-3p and miR-373-3p both in mouse plasma and tumour tissue samples are correlated with the development of tumour-containing EC elements. To determine whether the presence of a humanised immune system in the mouse would affect the efficiency of these miRNAs as predictive markers of malignancy, we analysed bi-weekly and endpoint plasma and tumour samples.
The accumulation of circulating miRNAs in plasma correlated with tumour growth in NSG females injected with 2102Ep and LU07+dox, both for T1 and T2 tumours (2102Ep only). Furthermore, the build-up of these miRNAs preceded the first detection of the tumour ( Figures 4A and S5A). LU07+dox hiPSCs also generated tumours containing derivatives of the three germ layers ( Figure 3B) but additionally contained areas with EC components characterised by the expression of OCT4 and CD30 ( Figure 3D). The presence of EC components was con sistent in all NSG males, NSG, and HIS females. We examined sections of the lymph node (subiliac and mesenteric, liver, lungs) by H&E staining and found no signs of metastasi in any of the organs (data not shown).
Teratomas tend to be cystic tumours, in which cysts containing fluid are in between solid areas ( Figure S4). Possibly, the cystic structures hamper the determination of th actual tumour mass. Therefore, the solid surface areas in tumours generated by LU07 and LU07+dox in NSG male and females and in HIS female mice were determined in seria are correlated with the development of tumour-containing EC elements. To determine whether the presence of a humanised immune system in the mouse would affect the efficiency of these miRNAs as predictive markers of malignancy, we analysed bi-weekly and endpoint plasma and tumour samples.
The accumulation of circulating miRNAs in plasma correlated with tumour growth in NSG females injected with 2102Ep and LU07+dox, both for T1 and T2 tumours (2102Ep only). Furthermore, the build-up of these miRNAs preceded the first detection of the tumour ( Figure 4A, Figure S5A).  Similar expression patterns were observed for HIS mice, where a high expression of miR-371a-3p and miR-373-3p also preceded the development of tumour-containing EC components (LU07+dox and 2102Ep). In addition, these miRNAs were not detected for HIS mice injected with LU07 ( Figures 4B and S5B). NSG males injected with 2102Ep showed a similar pattern of accumulation of circulating miR-371a-3p and miR-373-3p preceding tumour development ( Figure S5C), in line with our previous report [13]. Once the tumours were removed (T1), the levels of miRNAs were undetectable in LU07-and LU07+dox-injected mice. In T1 tumour samples, higher expression of miR-371a-3p and miR-373-3p also indicated the malignant nature of the injected cell line, as levels of both miRNAs in mice injected with 2102Ep and LU07+dox were significantly higher than those in LU07-derived xenografts. Nonetheless, no significant differences were found in miRNA levels among mouse groups injected with LU07, LU07+dox, or 2102Ep ( Figure 4C).
A single LU07 tumour in one HIS female (T186) presented increased levels of both miRNAs in plasma samples and tumour samples even though histological analysis did not reveal any EC component. This particular tumour had rare, small pockets of OCT4expressing cells, which did not have EC morphology and did not express CD30 (data not shown), and therefore, were classified as undifferentiated cells.

Macrophages Are Present in the Fibrous Tissue Surrounding the Tumour but Rarely in the Tumour Parenchyma
Since HIS mice contain functional human immune cells, we examined the presence of F4-80 positive macrophages in the hPSC-derived tumours. In all animals, macrophages were present in the tissue surrounding the tumour (fibrous capsule) in varying amounts ( Figure 5A,B top panels), independent of the cell line or mouse model. Therefore, the presence of macrophages in tumour parenchyma was examined. No F4-80 positive macrophages were detected in the parenchyma of 2102Ep-derived tumours (data not shown). In tumours generated in NSG males from LU07 hiPSCs, macrophages were sparsely present in the solid part (non-cystic part) of the tumours. Similar results were obtained in NSG females and HIS females, with few F4-80 positive cells in areas of the tumours. In LU07+dox-generated tumours, macrophages were rarely detectable in compact tumour areas. We quantified the number of F4-80 positive cells per mm 2 of the tissue in serial sections of the tumour sections and indeed HIS females presented significantly higher cell numbers in LU07+dox tumours when compared with NSG males and females ( Figure 5C).

Discussion
Humanised mouse models have been used in various areas of immunology, including allergy, autoimmunity, infectious disease, and cancer [19] but, to our knowledge, not to evaluate the malignancy of stem cells in the in vivo teratoma assay. At the actual advancement of assay development, the teratoma assay remains relevant since it is the only assay to provide an assessment of pluripotency and malignant potential, which are both

Discussion
Humanised mouse models have been used in various areas of immunology, including allergy, autoimmunity, infectious disease, and cancer [19] but, to our knowledge, not to evaluate the malignancy of stem cells in the in vivo teratoma assay. At the actual advancement of assay development, the teratoma assay remains relevant since it is the only assay to provide an assessment of pluripotency and malignant potential, which are both relevant to the preclinical safety assessment of hPSCs [5].
In a side-by-side comparison, we analysed several hPSC lines with distinct differentiation capacities in the teratoma assay performed in NSG and HIS mice. We compared the time of first appearance and the dynamics of tumour growth generated from benign (LU07) and malignant (LU07+dox and 2102Ep) cell lines and observed no significant differences between NSG and HIS mice. Furthermore, the variations between the cell lines we observed reflected the nature of the cell line (fast-growing 2102Ep vs. slow-growing LU07 and LU07+dox) and were similar in NSG and HIS mice. Histopathological features were similar in HIS and NSG mice and were consistent with our previous data [13]. LU07 hiPSCs all formed teratomas containing the three germ layers and were lacking undifferentiated areas, as shown by H&E and OCT4 staining. By contrast, LU07+dox cells, previously reported to be malignant [13,18], generated teratomas composed of tissues derived from three germ layers but with the presence of EC components, which closely resembled hECs generated from 2102Ep cells, expressing both CD30 and OCT4 pluripotency markers.
Minimally invasive, liquid biopsies are already used for the diagnosis of hGCTS [20,21]. Using this approach, the cell lines used here (LU07, LU07+dox, 2102Ep) were previously miRNA-profiled after injection into NSG mice [13]. Here, we used the same strategy, by generating tumours using NSG mice as a control to determine whether HIS mice were sensitive enough for malignant miRNA detection. We examined mouse plasma miRNA levels after injection of hPSC at the endpoint of the xenograft and compared these levels with those before injection. We analysed miRNA371a-3p and miRNA373-3p expressions, previously reported to be good predictive markers for malignancy in hGCTs [12,22] and hPSCs [13]. The expressions of these miRNAs in 2102Ep tumours increased before the tumours were visible and decreased after surgical tumour removal. In malignant LU07+doxinjected animals, increased levels of miRNA371a-3p and miRNA373-3p were observed in NSG and HIS mice, which all contained EC components. We confirmed this finding also in HIS mice: miRNAs patterns reflect the histological constitution of the tumours even before tumours are measurable by calliper [13]. We also showed that, after surgical removal of tumours (T1 in HIS and NSG mice), miRNAs are cleared from the circulation, as has also been observed in patients after surgical removal of clinically manifested hGCTs [22,23]. The patterns found in the mouse xenograft models mimic those in hGCT patients. Interestingly, the miR-371 family has been found to be an alternative mechanism for the inactivation of the P53 pathway in hGCTs [24]. TP53 mutations have already been reported in human embryonic stem cells, suggesting that these mutations confer a selective advantage. After sequencing the protein-coding genes (exomes) of 140 independent hESC lines, Merkle et al. concluded that acquisition and expansion of cancer-associated mutations in hPSCs may go unnoticed during most applications and advised careful genetic characterisation of hPSCs and their differentiated derivatives prior to clinical use [25].
Combined, the data presented here suggest that the existence of a partially reconstituted human immune system is not advantageous in a teratoma assay over NSG mice for detecting malignancy of stem cells. In clinical situations, it is more likely that the immune system will be, at least partially, active. Macrophages as a part of the innate immune system are the first line of defence for the organism against pathogens and invading foreign or transformed cells. Generally, M1 macrophages (classically activated macrophages) are responsible for inflammatory response [26,27], while M2 macrophages (alternatively activated macrophages) for parasite infection, tissue remodelling and angiogenesis [28,29]. Once associated with a tumour tissue, tumour-associated macrophages (TAMs) can affect tumour progression and either have tumour-promoting effects [30] or exhibit antineoplas-tic activity [31]. Therefore, functional macrophages could influence the outcome of the assay. However, the assessment of macrophage infiltration in the tumour in NSG and HIS mice did not reveal major differences between animals. HIS females injected with malignant LU07+dox cells exhibited increased numbers of macrophages, compared with those injected with benign LU07 cells, which might be a consequence of the presence of human macrophages. Alternatively, since TAMs can have the properties of either M1 or M2 macrophages, tumours in HIS mice underwent more tissue remodelling, thereby attracting more macrophages. Interestingly, there was a significant difference between the numbers of F4-80-positive macrophages in NSG males and females injected with LU07. However, since the macrophages in NSG mice are defective [17], this would not affect the immune response.
In conclusion, when a teratoma assay is performed to detect malignancy of a cell line, NSG mice appear equally sensitive as HIS mice.
The teratoma assay raises both ethical and methodological questions. It is performed on mice in which tumours of considerable size need to develop, creating obvious animal welfare issues. Moreover, it is a time-consuming assay that requires expert pathological assessment, which is difficult to quantify and impossible to apply as a routine and largescale screening tool [32]. Importantly, despite the several calls [6,33], the teratoma assay has never been standardised. Additionally, the generation of HIS mice requires irradiation of newborn animals, adding to animal discomfort [17].
Future research is urgently needed to replace the teratoma assay with in vitro methods based on genetic and epigenetic biomarkers able to identify cell lines with malignant potential.

Cell Lines
Human embryonal carcinoma 2102Ep [34,35] and human-induced pluripotent cells LUMC007iCTRL01 (LU07) were used in this study. The generation of the LU07 cell line with the doxycycline-inducible TetO-FUW-OSKM construct was described previously [18]. Activation of the transgenes results in continuous expression of reprogramming factors (OCT3/4, SOX2, KLF4, MYC), rendering the cell line differentiation defective. LU07 cells were cultured on VitronectinXF (Stem Cell Technologies)-coated, standard tissue culture plates in mTeSR-E8 (Stem Cell Technologies) culture medium, refreshed daily, and passaged weekly as clumps with a Gentle Cell Dissociation Reagent (GCDR, Stem Cell Technologies); 2102Ep cells were cultured as described [18]. The LU07 cell lines without activation of the transgenes are abbreviated in this manuscript as LU07. Expression of the transgenes was activated and maintained by the addition of doxycycline (2 µg/mL final concentration) in culture media for 3 days prior to use for injections into animals. These cells are abbreviated as LU07+dox [18].

FACS Analysis
On the day of cell injection, LU07 and LU07+dox cells were tested for pluripotency. Single-cell suspensions were processed with Fix&Perm Cell fixation and Permeabilisation Kit (Invitrogen), according to the manufacturer's instructions. Cells were incubated with anti-OCT3/4-Isoform A-PE antibody (Miltenyi) or isotype control IgG-PE (REA-PE, Miltenyi) and analysed using an LSRII analyser with Diva 8.02 software (BD). Cultures were used for injection only if >90% of the tested population expressed OCT3/4. Detailed information on antibodies used is presented in Table S1.

Animals and Teratoma Assay
Mice used in this study were 18-21-week-old NOD, Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) males, 22-week-old NSG females, and 25-week-old NSG females engrafted with human CD34+ hematopoietic stem cells (HIS mice). All animals were purchased from the Jackson Lab (USA) and, upon arrival, were housed in sterile conditions in individually ventilated cages. All experiments were performed at Leiden University Medical Centre (LUMC) and were approved by the Dutch Central Commission for Animal experimentation (permit: ADV116002016735) and performed according to ARRIVE guidelines [36]. Group allocation is presented in Figure S1. Briefly, 1 × 10 6 cells in 200 µL Matrigel mixed with cold mTeSR-E8 1:1 (Corning/Gibco) for hiPSCs and DMEM-F12 for 2102Ep were subcutaneously injected into the mouse's right flank. This time point is referred to as T0. Animals receiving LU07+dox cells were given 2 mg/mL doxycycline (Sigma-Aldrich), with 10 mg/mL sucrose (Sigma-Aldrich) in drinking water to maintain construct expression, while animals receiving LU07 cells were provided with only sucrose in drinking water [18]. Water with additives was changed every other day.

Plasma Sample Collection for mi-RNA Analysis
For each animal, plasma collection with tail bleeding was performed before cell injection (T-1) one week after injection, followed by bi-weekly collection throughout the entire experiment. Additional sample collection was performed at tumour removal if the previous collection was more than 7 days earlier. Blood was collected to heparin-containing collection tubes and centrifuged at 5000 rpm for 5 min; plasma was removed and stored at −80 • C until miRNA isolation and analysis.

Tumour Growth Monitoring
Tumour growth was monitored weekly by palpation and measured with a digital calliper. The tumour volume (V) was calculated according to the formulation V = (Wˆ2*L)/2 [37], where W is width, and L is length.
Once the tumour reached~2 cm 3 (T1), it was surgically removed under anaesthesia and perioperative analgesia. T1 tumours were processed for histological analysis. The animal was kept alive up to 8 weeks after T1 removal and closely monitored for tumour growth and animal welfare. At the experimental endpoint (T2), the animals were sacrificed, and internal organs-and, if present, the tumour (T2)-were collected and, after fixation, embedded in paraffin for analysis.

Tumour Histology
Tumour tissue was fixed in 4%PFA (Sigma-Aldrich) overnight at room temperature (RT) and embedded in paraffin. Paraffin sections (5µm thick) were prepared according to standard protocols and stained with haematoxylin and eosin (H&E) for morphological analysis, as described [18]. Evaluation based on H&E stain was performed by a European board-certified veterinary pathologist (D.C.F.S) and by one of the authors (M.B.). For immunohistochemical staining with various antibodies (Table S1), antigen retrieval was performed by heating the slides at 95 • C for 12 min either in citrate buffer pH 6.0 or Tris-EDTA buffer pH 9.0. Sections were permeabilised with 0.05%Tween or 0.1%Triton, blocked in 0.05%Tween-1%BSA for 1 h at RT, and incubated with primary antibodies in a blocking solution overnight at RT. Appropriate biotinylated secondary antibodies were diluted in a blocking solution and incubated for 1.5 h at RT. Signal amplification was performed by incubation with ABC-HRP Kit (Vector Laboratories) reagent, followed by DAB detection (Vector Laboratories) or BrightDAB (ImmunoLogic). Nuclei were counterstained with haematoxylin. A list of used antibodies is provided (Table S1).

RNA Isolation and miRNA Analysis
miRNA from plasma samples and tumours was isolated as described [13]. For additional analysis of miRNAs, tumour sections were deparaffinised and dehydrated following standard procedures. Tissue was scraped off the glass slides and collected in TRIzol™ reagent (Invitrogen). Total RNA and miRNAs were retrieved using a Direct-zol™ RNA MicroPrep Kit (Zymo Research), according to the manufacturer's protocol. miRNA levels were normalised using the endogenous control RNU48 [38].

Imaging and Quantification
Images were taken with Panoramic Scanner 250 (3DHISTECH) and CaseViewer 2.4 software. Brightness/contrast adjustments were made with Fiji software [39]. Figures and illustrations were assembled in Adobe Illustrator CS5.

Volume Calculation from Sections
To calculate tissue volume from sectioned material, we used all available serial sections (sectioned 1:10) stained with H&E of each tumour. Tissue area was measured on every slide using the 'measure particles' tool in Fiji software, multiplied by the thickness of the section (0.005 mm) and the distance between sections (0.05 mm). Since tumours varied in size and the number of sections obtained, the summarised values for each tumour were divided by the number of sections used for counting. The amount of sections used for each tumour was T182 n = 23; T183 n = 17; T184 n = 19; T185 n = 13; T186 n = 7; T187 n = 21; T188 n = 12; T189 n = 20; T190 n = 17; T191 n = 19; T192 n = 10; T193 n = 11; T194 n = 13; T195 n = 16; T196 n = 18; T197 n = 12; T198 n = 20; T199 n = 11.

Macrophage Counting
Selected slides from LU07, LU07+dox, and 2102Ep generated tumours were stained with macrophage marker F4-80 using mouse lung and spleen as positive controls. Positive cells were manually counted on the entire slide in the tumour parenchyma using Case-Viewer software. The area of the section (in mm 2 ) was calculated in Fiji using the 'analyse particles' option after threshold adjustment. Areas of the tumour at the edges which were not derived from hiPSCs (judged based on anti-human nucleus stain) were excluded from cell count and area measurement.

Statistical Analysis
All graphs and statistical analysis (2-way ANOVA test) were made in Prism 6.0c. Statistical analysis for macrophage counting was performed using the Mann-Whitney test.