This research has been performed in accordance with the Declaration of Helsinki and was approved by the Ethics, Research and Biosafety Committee from IMIEM (CIEICE-007-01-13) and by the National Committee of Scientific Research at IMSS (R-2012-3602-29 and R-2015-785-120). All samples were collected after informed consent from parents. The study included 147 B-ALL pediatric patients, 8 months to 16 years old (8.15 ± 4.47), referred to the IMSS Specialties Hospital and the IMIEM Children’s Hospital. At clinical diagnosis, 85% of patients were classified as high risk and 41.5% as ProB/PreB-, 30.6% as ProB-, and 27.9% as PreB-ALL, with only 30% exhibiting prognostic translocations. Control BM was obtained from 12 healthy children undergoing minor orthopedic surgery. BM specimens were collected by aspiration before any treatment and according to international and institutional guidelines. ( Supplementary Table S1 ).

Mononuclear cells (MNCs) were separated by Ficoll-Paque Plus (GE Healthcare Bioscience, NJ, USA) gradient. No sample pooling was performed for any of the experimental strategies. By immunophenotyping with fluorochrome-conjugated antibodies ( Supplementary Table S2 ), Pro-B cells were identified as CD45^low/-CD34⁺CD10⁺CD19⁺ and Pre-B as CD45^low/-CD34^-CD10⁺CD19⁺, before sorting in a FACSAria II flow cytometer (BD Biosciences, USA) ( Supplementary Figures S1A, B ).

REH and RS4;11 B-ALL cell lines were purchased from ATCC (VA, USA) and cultured according to instructions. Nalm6 cell line was kindly provided by Dr. JL Maravillas (INNSZ, Mexico). Cell lines were tested for mycoplasma and authenticated using STR assays.

MSCs were isolated by adhesion, as previously reported (22).

A total of 25,000 MSCs were plated on 96-well round-bottom plates previously coated with 1% agarose to induce spheroid formation for 24 h, before co-culture with leukemic cells (22). For harvesting, PDLS were incubated with 0.05 mM PBS-EDTA for 5 min to detach cells from the surface, followed by 10 min enzymatic treatment (TrypLE Express, Gibco, CA, USA) and mechanical disruption. Cell suspension was recovered from the inside of PDLS (PDLS-in) and separated from outer cells and supernatant (PDLS-out), before staining with fluorochrome-conjugated antibodies and/or direct FACS analysis ( Supplementary Figures S1C, D ).

MSCs or B-ALL cells were stained with fluorescent dyes Cell Trace Violet^®, Cell Trace CFSE^®, or Cell Trace Far Red^® (Invitrogen, Life Technologies, CA, USA), according to the manufacturer.

PDLS were fixed with 4% PFA and 2 h treated with 0.01% Triton X-100 (Bio-Rad, MX). Upon 1 h blocking with 3% BSA, they were incubated overnight at 4°C with primary unlabeled antibodies in PBS 3% SFB, washed, and incubated for 1 h with conjugated secondary antibody before 10 min DAPI staining and Vectashield. BM biopsy staining was performed as described (22).

Supernatants were collected after 24 h of 3D culture and investigated for cytokines by multiplex assays (Milliplex Map, Millipore, Merck MX).

FACS-sorted B-ALL cells were stained with CellTrace CFSE^® (Invitrogen, Life Technologies, CA, USA), co-cultured with MSC spheroids, and further assayed for fluorescent dye dilution by flow cytometry.

Hypoxia was detected by the Hypoxyprobe-1 Plus Kit (Pimonidazole Hydrochloride, Chemicon International, Temecula, CA, USA). Pimonidazole incorporation was confirmed by flow cytometry and fluorescence microscopy. Image-IT green hypoxia (Invitrogen, Life Technologies, CA, USA) was used to track low oxygen levels.

Harvested cells were adjusted to 10⁶ cells/ml and incubated with Hoechst 33342 to a final concentration of 5 μg/ml (Sigma-Aldrich, MX), 37°C for 2 h, prior to staining with anti-human CD45.

In vivo experiments were conducted according to the WCM Institutional Animal Care and Use Committee (IACUC) and the CINVESTAV Committee for Animal Care and Use (CICUAL) guidelines and regulations. NOD/SCID gamma chain (NSG) mice from the Jackson Laboratory (JAX, CA, USA) were i.v. injected with primary B-ALL cells from 48 h cultures. Animals were euthanized after 5 weeks or when exhibiting clinical signs of leukemic disease. Human CD45⁺ cell frequencies in peripheral blood and BM were investigated for engraftment monitoring.

Serial dilutions of leukemic cells were injected into NSG mice. After 4 weeks, the engraftment was determined by flow cytometry and documented as positive when human CD45⁺ cells recorded within mouse BM cells were >1%. ELDA program was used to calculate the LIC content for each culture condition (23).

Whole RNA was extracted from 5 × 10⁵ MSC (RNeasy kit, QIAGEN, MX), and samples with RIN > 8 were used for experiments. Libraries were constructed by using the TruSeq Stranded mRNA Library Prep Kit (Illumina, CA, USA) before mRNA sequencing on a NextSeq 500 instrument at INMEGEN (Mexico). Paired-end reads were aligned to the human genome reference GRCh38/hg38 (build 38.2) with the R software package Rsubread (24) and read mapping statistics were reported ( Supplementary Table S3 ). Mapped reads were summarized to gene level counts featured by counts function of Rsubread, considering the built-in NCBI RefSeq gene annotation for gene reference. Protein coding genes with detected counts in at least one sample library were retained and normalized using TMM normalization. Differential expression analysis was performed with the edgeR package (25). Statistical analyses and plots were performed using the programming language R (R Core Team, 2012). Gene ontology and functional enrichment analyses were performed by Metascape (26). The original contributions presented in the study are publicly available. RNA-seq data can be found in E-MTAB-10838 (https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-10838/).

FlowJo 10.0.8 (TreeStar Inc., Ashland, OR, USA) and Infinicyt 1.8 (Cytognos, Spain) software were used for cytometry data, while Prism 8 (GraphPad, CA, USA) software was used for statistical analysis. Differences within groups were established by non-parametric tests, considering significant probability values <0.05. Mann–Whitney U test with α of 5% to define significance was applied. Data were normally distributed and individual data points for independently repeated experiments and mean (SD) were graphed.

As 3D cellular organization is essential to preserve physiological features of BM, we generated 3D structures to characterize the leukemic niche by using MSCs derived from either primary B-ALL at disease onset (ALL-MSC) or from healthy bone marrow donors (HBM-MSC) ( Figure 1A ). We found that MSCs were capable of forming a single multicellular spheroid within the first 24 h of non-adherent culture conditions with a direct cell number–size relationship ( Supplementary Figures S2A, B ). Despite the decrease in the MSC proliferation in 3D settings, the classical MSC markers were conserved ( Supplementary Figures S2C, D ). CXCL12-abundant reticular (CAR) immunophenotype was confirmed in CXCL12^hiSCF^hiIL-7^hi HBM-MSC spheroids ( Figure 1B ) as well as the expression of Nestin, PDGFRα, and LepR ( Supplementary Figure S2E ). In contrast, a substantially lower abundance of typical CXCL12^hiSCF^hi CAR cells in ALL-MSC spheroids with weaker expression of CXCL12 and SCF ( Figure 1B ), but increased production of IL-8, Flt3-L, GM-CSF, FGF-2, CXCL10, and CXCL11, was observed in the supernatants evaluated at 24 h ( Figure 1C ). ALL-MSCs in a 3D organization have the ability to in vitro recapitulate unique CAR niche-like structures that produce high levels of CXCL10 and CXCL11.

Since we confirmed the ability of the 3D ALL-MSCs spheroids of recapitulate CAR niche-like structures, we sought to evaluate the ability of primary B-ALL cells to migrate to the MSC spheroids by assessing their colonization capacity. First, we stablished the 3D HBM- or ALL-MSCs spheroids, and 24 h later, we seeded 25,000 primary B-ALL cells CD10⁺CD19⁺ (n = 8), labeled with Cell Trace Far Red. After 24 h of co-culture, spheroids were washed and prepared for the whole-mount fluorescence microscopy analysis or enzymatically disrupted to analyze their cellular content by multiparameter flow cytometry (MPFC). We found a clear advantage for ALL-MSC spheroids to facilitate the colonization of leukemic cells when compared with the HBM-MSC ( Figure 2A ). To evaluate niche saturation, serial spheroid sizes were tested, finding that in all cases, only near 1%–3% of leukemic cells were able to colonize inner niches ( Supplementary Figure S3A ). Since CXCR4 has been implicated in homing of leukemia cells to their niche (27, 28), we tested the effect of plerixafor (AMD3100) in the colonization of AMD3100-treated B-ALL cells to the 3D structures and found that it can only partially prevent B-ALL cell spheroid colonizing, with a 4.6-fold decrease ( Figure 2B ). The effect was similar when the niche positioning of leukemic cells into normal BM spheroids was investigated ( Supplementary Figure S3B ).

Next, we assessed the ability of the spheroids to maintain cells capable of initiating leukemia without enrichment. Thus, we co-cultured 25,000 MNCs from five different B-ALL patients with either stromal-free (SF), MSCs monolayers (2D), or spheroids (3D) for 48 h and then transplanted into NSG mice ( Figure 2C ). We observed that 3D architecture was best at facilitating survival and expansion of primary leukemia cells when compared to other culture conditions ( Figure 2D and Supplementary Figure S3C ). Importantly, 3D co-cultures in ALL-MSC expanded more robustly ( Figure 2D ) and exhibited higher leukemic engraftment at week 6 post-transplantation than other culture conditions ( Figure 2D and Supplementary Figure S3D ). In addition, cells cultured in the 3D system performed better than freshly thawed MNC and transplanted ( Supplementary Figure S3E ). Taken together, we demonstrated that ALL-MSC spheroids support homing, survival, growth, and efficient engraftment of primary B-ALL cells. This co-culture system is referred to as patient-derived leukemic spheroids (PDLS).

Despite the fact that LICs in B-ALL have been controversial due to the lack of a specific immunophenotype (18), cells with stem cell features have been shown to be enriched in hypoxic zones within the BM (29). Here, we sought to characterize the cells capable of colonizing the PDLS. Because leukemia initiation in NSG mice is a feature of LICs, this and additional LIC properties were evaluated in different compartments of 3D structures using primary B-ALL samples. At 24 h, we harvested cells from the supernatant (PDLS-out) and, upon enzymatic digestion of the PDLS structure, collected the cells that migrated into the inner spheroid (PDLS-in). More than 90% of spheroid-colonizing B-ALL cells (PDLS-in) showed low proliferation activity when growing inside PDLS, while PDLS-out cells exhibited higher proliferation ( Figure 3A ). Consistently, a quiescent (G0) profile defined the PDLS-in cells ( Figure 3B ), which is a feature of LICs (18). Furthermore, we investigated stem cell features such as “side population” ( Figure 3C and Supplementary Figure S4A ) and low ROS production (30) ( Figure 3D ), confirming that PDLS-in cells also displayed such properties when compared with other culture scenarios. Moreover, an increase in HIF-1α expression was recorded ( Figure 3E and Supplementary Figure S4B ), consistent with increased hypoxia in the PDLS-in cells, assessed by the image-iT green hypoxia tracker and pimonidazole incorporation. These data confirmed a PDLS-in hypoxic setting for both MSCs and B-ALL ( Figure 3E and Supplementary Figure S4C ). Taken together, PDLS provide strong evidence that stem-like B-ALL can be enriched by their function and biological features within hypoxic niches, suggesting that they may be the foundation of leukemia-migrating and -proliferating cells.

As PDLS were colonized by leukemia cells with stem cell features, we sought to determine whether cells isolated from PDLS-in are characterized by the increased ability of homing. By serial spheroid seeding assay, we discovered that PDLS-in were capable of re-colonizing spheroids with higher efficiency than PDLS-out cells ( Figure 4A ), highlighting their homing and stem cell potentials. To further characterize the LICs capacity, 3,000 sorted CD45⁺ RS4;11 cells from 48 h PDLS-in and other culture scenarios were used to inject NSG mice. Leukemia burden was weekly monitored, and final engraftment was evaluated at 6 weeks ( Figure 4B ). Mice transplanted with purified PDLS-in cells showed the highest numbers of human CD45⁺ cells peripheral blood (PB) and exhibited the lowest overall survival (OS) of 44 days ( Figure 4C ). BM analysis confirmed the facilitated engraftment with PDLS-in RS4;11 cells. Such results were validated with three different primary human samples from ProB and PreB pediatric-ALL patients ( Figure 4D ). Limiting dilution assay revealed that LICs frequency was 10 times less in stroma-free settings when compared to PDLS-in conditions ( Supplementary Figures S5A, B ). Remarkably, a LICs enrichment was observed in PDLS-in (1/45.2), compared with PDLS-out (1/858) and SF conditions (1/704) ( Supplementary Figure S5C ). MPFC analysis of PDLS-in confirmed that LICs enrichment by PDLS was likely driven by functional attributes associated with leukemia stemness rather than by immunophenotype ( Supplementary Figure S5D ), supporting the notion of a functional LICs hierarchy driven by specialized microenvironmental cues. Thus, PDLS could be potentially used as a proxy to determine the presence of LICs in pediatric B-ALL patient samples.

As LICs have also been described as chemo-resistant (13, 18, 27, 31), we proceeded to investigate the response of the PDLS-in cells to the most commonly used chemotherapy drugs for B-ALL treatment. To this end, the ability of drugs to diffuse inside the spheroid was investigated. When treating PDLS with the anthracycline daunorubicin, the cells were able to uptake daunorubicin within the first hour, evidenced by their red fluorescence ( Figure 5A ). By examining the viability at 24 h of treatment, we found that daunorubicin, prednisolone, and vincristine, even at high concentrations, were not effective in killing the PDLS-in cells ( Figure 5B ). Of note, combined chemotherapy commonly used in B-ALL, including daunorubicin, prednisolone, vincristine, and methotrexate (P-V-D-M), displayed similar results when investigated in high-risk (HR) and standard-risk (SR) patients ( Figure 5C ). Furthermore, when PDLS-in vehicle or P-V-D-M-treated cells were purified and exposed for an additional 24 h in stroma-free conditions, cells remained chemo-resistant ( Figure 5D ). Next, to determine the potential of PDLS-in cells to recapitulate disease after chemotherapy, PDLS were treated with combined chemotherapy for 24 h, washed to remove PDLS-out cells, and cultured again in fresh wells. Strikingly, newly formed PDLS-out cells were harvested upon 120 h and no differences were observed when compared to untreated PDLS ( Figure 5E ), suggesting that PDLS can capture clinical features, such as tumor reemergence after cell survival within internal niches during chemotherapy.

In order to investigate the identity of MSCs isolated from primary pediatric-ALL patients, we performed RNA sequencing analysis of three different ALL-MSCs specimens and HBM-MSC. Substantial and heterogeneous dysregulation of gene expression was found when compared with HBM-MSC ( Figure 6A and Supplementary Table S3 ). Specifically, 103 genes were consistently overexpressed among ALL-MSC (fold change > 2 and FDR < 0.05) ( Figures 6B, C and Supplementary Figures S6A, B ) and, of high interest, two major gene ontology (GO) signatures were identified. A pro-inflammatory signature was characterized by a large set of chemokines involved in neutrophil recruitment, IL-17 signaling, metalloproteinase functional activation, and leukocyte migration including CXCL1, CXCL2, CXCL3, CXCL5, CXCL6, CXCL8, CCL20, and pro-inflammatory molecules like IL1B, IGF1, MMP1, MMP3, and MMP8 ( Figure 6D ). An additional signature, predominantly displayed by ALL-MSC3, showed a TLR signaling, cytokine-mediated signaling, and a negative regulation of leukocyte proliferation signatures. Moreover, high expression of chemokines CXCL10 and CXCL11 and a substantial expression of suppressor molecules like indoleamine 2,3-dioxygenase (IDO1) and galectin 9 (LGALS9) ( Figure 6D ) were apparent. Importantly, ALL-MSC did not exhibit transcriptional differences in the typical MSCs markers CD73, CD90, and CD105 ( Supplementary Figure S6C ), but a very low transcriptional expression of CAR-niche associated genes CXCL12 and SCF were found in CXCL10⁺CXCL11⁺ ALL-MSC3 ( Supplementary Figure S6D ). When downregulated genes were analyzed, we did not find apparent intersections among samples. However, GO analysis at the individual level showed that some extracellular matrix-associated proteins and cell division-associated networks were dramatically altered in similar extent for all three B-ALL MSCs ( Supplementary Figure S6A ). Taken together, the MSC gene expression profiling suggests two potential niches, according to their functional elements within the B-ALL BM microenvironment. A pro-inflammatory and leukemia expansion (ILE) niche, where leukemic clones may proliferate and increase tumor burden in the context of an activating pro-inflammatory milieu, and an immune-suppressive and leukemia-initiating cell (SLIC) niche, endowed with immunoregulatory and suppressive properties and high transcription of CXCL10 and CXCL11.

In order to assess our transcriptional observations in our PDLS system, we used immunostaining approaches to characterize CXCL11 expression in a leukemia microenvironment. Strikingly, we found that CXCL11^hi MSC spheroids were enriched in hypoxic CXCL12^hi zones with partial overlapping ( Figure 7A ). Distinct CXCL11^low and CXCL11^hi cell populations were also evident in ALL-MSC spheroids, while HBM-MSC spheroids did not show CXCL11 expression ( Figure 7B ). Moreover, the occurrence of CXCL11^low/hi MSCs in B-ALL BM biopsies was confirmed ( Figure 7C ), where CXCL11 co-stained with CD19. Additionally, B-ALL cells, but not normal CD34⁺ precursor cells, expressed CXCR3 and CXCR7, the receptors for CXCL10, CXCL11, and CXCL12, suggesting their advantage for selective niche colonization ( Figures 7D, E and Supplementary Figure S7A ). CXCL10^hiCXCL11^hi zones may represent exclusive leukemia-positioning niches where B-ALL cells may also contribute to CXCL11 expression ( Supplementary Figure S7B ) presumably relevant for positioning of CXCR3⁺CXCR7⁺ LICs and suitable for immune scape ( Figure 8 and Supplementary Figure S7C ). Taken together, we demonstrate that PDLS are capable of capturing a SLIC niche endowed with a specialized gene expression signature and the selection of malignant cells with stem cell functions.