SEARCH

SEARCH BY CITATION

Keywords:

  • Hematopoietic stem cells;
  • In vivo tracking;
  • Hematopoiesis;
  • Lentiviral vector

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Our understanding of system dynamics of mixed cell populations in whole organisms has benefited from the advent of individual cell marking by nonarrayed DNA barcodes subsequently analyzed by high-throughput DNA sequencing. However, key limitations include statistical biases compromising quantification and the lack of applicability to deconvolute individual cell fate in vivo after pooling single cells differentially exposed to different conditions ex vivo. Here, we have derived an arrayed lentiviral library of DNA barcodes and obtained a proof-of-concept of its resolving capacity by quantifying hematopoietic regeneration after engraftment of mice with genetically modified autologous cells. This method has helped clarify and bridge the seemingly opposed clonal-succession and continuous-recruitment models of hematopoietic stem cell behavior and revealed that myeloid-lymphoid biases are common occurrences in steady-state hematopoiesis. Arrayed lentiviral barcoding should prove a versatile and powerful approach to deconvolute cell dynamics in vivo with applications in hematology, embryology, and cancer biology. STEM Cells 2013;31:2162–2171


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Hematopoietic stem cells (HSCs) are the best characterized among postnatal cells capable of self-renewal and further differentiation, and HSCs serve as a paradigm for the study of other stem cell populations. Because they can reconstitute the entire hematopoietic system after transplantation, HSCs are routinely used in cell-based therapies to treat numerous hematologic [1] and nonhematologic [2] diseases in allogeneic or autologous settings. In addition, ex vivo gene transfer to HSCs followed by autologous transplantation has paved the way to the field of gene therapy with proof-of-principle of efficacy recently obtained in human clinical trials [[3-6]]. However, improving our knowledge of HSC biology is highly desirable to clarify unresolved issues regarding their behavior and underlying molecular mechanisms. There is also an unmet need to increase HSC availability to optimize their utilization in regenerative medicine.

The quantification of HSCs, their manipulation in vitro, and progress in our understanding of their individual fate in vivo remain largely dependent on cumbersome mouse transplantation assays with limited number of input cells. These methods are costly, time-consuming, and ethically censurable as they require large numbers of mice. Moreover, they do not allow for competition experiments with different HSC populations in a unique transplant recipient, although this approach would provide precious novel information should it be technically feasible.

HSC tagging with DNA barcodes (BCs) integrated in the cell's chromosomes has been used to track different cell types simultaneously in vivo [[7, 8]] and was recently applied to replace single-cell analysis of HSCs [[9, 10]]. However, the libraries of BCs used in these studies were not arrayed. Hence, linkage of BCs and target cells could not be a priori determined, and complex statistical analyses were required to interpret the data, thus carrying a substantial risk of biases. In addition, nonarrayed BCs are not suitable for the deconvolution, within a pool of HSCs, of differential in vivo properties acquired after ex vivo exposure to a library of small molecules or other conditions.

Here, we have developed an arrayed lentiviral vector library containing BCs that can be effectively analyzed and deconvoluted after transplantation of mixed HSC populations in syngeneic murine hosts. The method was validated in mouse transplant studies and further used to gather new insights into topical questions in HSC biology: the quantification of HSCs having myeloid-lymphoid biases and the dynamics of clonal HSC contribution to steady-state hematopoiesis. We also determined a range of linearity between a given HSC input dose and BC blood cell output as a prerequisite to the use of the method for in vivo deconvolution of arrayed small molecule libraries capable of producing differential HSC effects after transplantation.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Barcoded Lentivector Library Design and Preparation

Plasmid DNA was isolated from 50 bacterial clones of a library kindly provided by Dr. T. Schumacher (Division of Immunology and Central Microarray Facility, The Netherlands Cancer Institute, Amsterdam). BC sizes are between 150 and 555 bp long and made of semirandom stretches ((N8)-(SW)5)5-N8. BCs were inserted in the pLentilox3.4 vector. We arrayed this plasmid library for DNA sequencing of BCs and for subsequent lentivector (LV) library production. The 50 plasmids were sequenced with the same primer complementary to the 3′ end of the enhanced green fluorescent protein (eGFP) (5′-GTCCTGCTGGAGTTCGTGAC-3′).

Vector Production and Titration

Each LV carrying a given and sequenced BCn, referred to as LV3.4-BCn, was produced in the cell line HEK 293T. Briefly, 4 × 106 HEK 293-T-cells were seeded in Dulbecco's modified Eagle's medium with 10% heat-inactivated fetal bovine serum 48 hours before transfection by calcium phosphate precipitation with 13 μg of HPV275 (gag/pol expression plasmid), 3 μg of P633 (rev expression plasmid), 3.75 μg of psiN15 (Vesicular Stomatitis Virus glycoprotein G [VSV-G] expression plasmid), and 13 μg of LL3.4-BCn. Medium was replaced with StemSpan SFEM 6 hours post-transfection. Viral supernatants were collected 48 hours later, filtered through a 0.45 μm filters, and concentrated by ultracentrifugation at 4°C, 2 hours, 22,000 rpm. The viral pellet was resuspended to a final volume of 100 μl phosphate-buffered saline and stored at −80°C. Viral titers were estimated on HEK 293T and transduction efficiencies were confirmed on mouse bone marrow (BM) mononuclear cells. Briefly, 100,000 BM cells were seeded in 96-well plates, infected by various volumes of concentrated virus in the presence of 8 μg/μl protamine sulfate. After 8 hours, the medium was replaced with fresh growth medium. Transduction efficiency was determined 7 days later on eGFP expression evaluated by flow cytometry.

Cell Purification

BM cells were obtained by flushing the femurs and tibias of CD45.1 mice. Hematopoietic subpopulations were defined according to the expression of cell-surface markers and were analyzed and isolated by BD FACS Canto II and flow cytometry using BD FACS Aria, respectively. The Lin cell population was purified using the Lineage cell depletion kit (130-090-858, Miltenyi-Biotec, Paris, France, www.miltenyibiotec.com). The Lin population was further labeled with antibodies specific for Sca.1 (sca-APC 17-5981-81, eBioscience, Paris, France, www.ebioscience.com, diluted at 1/50) and Endoglin (endo-PE 130-092-929, Miltenyi Biotec, diluted 1/10). The HSC-enriched population was phenotypically identified as Lin Endoglin+ Sca.1+ (LES).

Cell Culture

LES cells were cultured together with total BM cells as growth support in 96-well plates in minimum essential medium alpha modification (α-MEM) supplemented with 15% of fetal bovine serum, 100 ng/ml of murine Stem Cell Factor, 10 ng/ml of murine IL-6, 6.25 ng/ml of murine IL-3, and 1% penicillin-streptomycin.

Cell Transduction

LES cells were preactivated overnight with the aforementioned cytokines and then infected with the different LV3.4-BCn at a multiplicity of infection (MOI) of 30 in the presence of 8 μg/μl protamine sulfate. After 48 hours of culture, cells were removed, pooled, and washed with α-MEM medium without any antibiotics or cytokines just before transplanting the cells into recipient mice.

Mice

C57BL/6J (CD45.2) and C57BL/6.SJL-Ptprc (CD45.1) were purchased from Charles River Laboratories (Les Oncins, France, http://www.criver.com). C57BL/6-tg(UBC-eGFP)30Scha/J (CD45.2, eGFP) mice expressing eGFP under the control of the human ubiquitin C promoter were obtained from Jackson Laboratories (Bar Harbor, Maine, www.jax.org). Mice were housed at the Commissariat à l'Energie Atomique (CEA) mouse facility (Fontenay-aux-Roses, France) and studies performed in accordance with French regulations and with consent of the local ethics committee.

Competitive Repopulation Assay

C57BL/6J-CD45.2 recipient mice were lethally irradiated 24 hours before transplantation in an IBL 137Cs γ-irradiator two times at 5.5 Gy/minute at an average rate of 1 Gy/minute. LV-barcoded cells were transplanted by intravenous route with 500,000 competitive BM cells obtained from a C57BL/6-tg(UBC-eGFP)30Scha/J (CD45.2, eGFP) mouse. Recipient mice were maintained on antibiotic water (neomycin 2 g/l) for at least 2 weeks after transplantation. At 8, 16, and 24 weeks after transplantation, blood samples were taken from the retro-orbital plexus. At 24 weeks post-transplantation, mice were also sacrificed to harvest BM cells. For serial transplantation, 5 × 106 total BM cells were extracted from each primary recipient 24 weeks post-transplantation and injected into irradiated secondary CD45.2 recipients.

Flow Cytometry

Before fluorescence-activated cell sorting (FACS) analysis and after red blood cells lysis, cells were stained with antibodies for donor and recipient CD45 allotypes (anti-CD45.1-APC Alexa Fluor 780, 47-0453 and anti-CD45.2-APC, 17-0454, each provided by eBioscience and diluted at 1/200) to quantify for the presence of each cell populations (CD45.1 eGFP derived from nontransduced donor cells, CD45.1 eGFP+ derived from LV-barcoded donor cells, CD45.2 eGFP derived from recipient mice, and CD45.2 eGFP+ derived from competitive BM cells). Viable cells (negative for 7AAD staining) were analyzed by BD FACS Canto II.

Purification of Lineage-Specific Cell Populations

Blood and BM cells from primary and secondary transplanted mice were collected and lineage populations were sorted based on phenotypical markers using BD FACS Aria. After red cells lysis, cells were labeled with B220-APC (17-0452, eBioscience, 1/400) and CD3-APC (17-0031, eBioscience, 1/400) antibodies for lymphoid markers, and Gr1-PE (12-5931, eBioscience, 1/1,000) and Mac1-PE (12-0112, eBioscience, 1/1,000) antibodies for myeloid markers. Sorted cells were frozen for genomic DNA extraction.

Genomic DNA Extraction

Genomic DNA was extracted from blood and BM samples using the blood-Nucleospin Blood kit (740951, Macherey-Nagel, Hoerdt, France, www.mn-net.com) according to the manufacturer protocol. Genomic DNA was quantified using the DNA quantification Kit from Bio-Rad (170-2480).

DNA BC Recovery and Identification by Gel Analysis

PCR-amplification of genomic DNA using primers surrounding the BCs (5′-TGCTGCCGTCAACTAGAACAC-3′ and 5′-GATCTCGAATCAGGCGCTTA-3′) was initiated with 40 ng of genomic DNA under the following conditions: 95°C for 5 minutes; 40 cycles: 95°C for 10 seconds, 65°C for 30 seconds, and 72°C for 50 seconds. This first PCR generates three sizes of amplicon easily detectable on a 2% agarose gel (555 bp and 360 bp bands containing one BC each, and 150 bp band containing two different BCs). Band quantification was performed with the ImageJ software. To take into account the PCR bias due to length and the GC percentage differences for each BC, we normalized results with the PCR yield for each BC. This was calculated using an equimolar pool of genomic DNA prepared from cells barcoded with independent BCs and a pool of genomic DNA prepared from individually barcoded cells at a ratio that exactly reproduces the BC composition established in vitro.

To calculate the percentage of each BC present in the 150 bp amplicon, we carried out another PCR in order to perform BC specific enzymatic digestion. This PCR step preferentially amplifies short DNA segments. For this purpose, PCR was performed with the forward and reverse primers 5′-GTCCTGCTGGAGTTCGTGAC-3′ and 5′-GCCATACGGGAAGCAATAGC-3′, respectively. The PCR was initiated with 40 ng of genomic DNA under the following conditions: 95°C for 5 minutes; 40 cycles: 95°C for 10 seconds, 59.4°C for 30 seconds, and 72°C for 50 seconds. Mixture of PCR products was digested separately by specific enzymes (BanII and AfeI) and the percentage of each BC among the 150 bp fragment was quantified. This percentage was reintroduced in the normalized results obtained after the first PCR to obtain the final percentage of each BC used.

DNA BC Recovery and Identification by Pyrosequencing

PCR amplification was performed with the forward and reverse primers surrounding the BCs (5′-TGCTGCCGTCAACTAGAACAC-3′ and 5′-GATCTCGAATCAGGCGCTTA-3′), extended with specific multiplex identifier (MID) sequences. Fourteen different MID sequences, each associated to one mouse, were used, allowing for pooling and simultaneous sequencing of 14 different samples. The PCR was initiated with 40 ng of genomic DNA under the following conditions: 95°C for 5 minutes; 40 cycles: 95°C for 10 seconds, 62.3°C for 30 seconds, and 72°C for 50 seconds. This first PCR generates the same three sizes of amplicon quantified with the ImageJ software. The 150 bp amplicons were simultaneously analyzed by sequencing on a sequencer 454 GS FLX Titanium from Roche. Different pools, composed of the same quantity of genomic DNA derived from cells labeled with only one BC, were sequenced to calculate the overall method yield for each BC. Sequences obtained were reassigned to each sample, and composition of each BC was calculated. This was corrected by the overall method yield and reintroduced in normalized results obtained after the first PCR to obtain the final percentage of each BC.

Statistical Analysis

Error bars were calculated using SEM. The frequency of HSCs in LES population after a 2-day culture was calculated using the L-CALC software (Stem Cell Technologies, Grenoble, France, www.stemcell.com) based on the reciprocal of the concentration of test cells that resulted in 37% negative mice. A given mouse was considered negative when flow cytometry analysis showed that it contained less than 1% of donor-derived cells within total peripheral white blood cells (WBC).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Experimental Strategy

The strategy is based on the introduction of a semirandom stretch of noncoding DNA (the barcode: BC) in HSCs using an LV (Fig. 1). The vector library is composed of BCs of 150–555 bp within an LV expressing eGFP under the control of the Ubiquitin C promoter. A set of 40 LV-encoding plasmids containing a different BC was sequenced, arrayed, and used to produce 40 BC LVs pseudotyped with VSV-G. Each LV titer was quantified on HEK 293T cells at limiting dilution by scoring the percentage of eGFP+ cells by flow cytometry 7 days after transduction. The results, presented as transduction units (TU) per milliliter, show that the transduction efficiency of the viral preparations is homogeneous and ranged from 3.6 × 106 to 1 × 107 TU/ml (Fig. 1). Titer homogeneity is essential to avoid misinterpretation by erroneously attributing differential transplantation output to what is a mere difference in ex vivo transduction efficiency.

image

Figure 1. Overall approach of the arrayed lentiviral BC strategy. The strategy is based on the introduction of a BC in hematopoietic stem cells (HSCs) using lentiviral vectors. The vector library is composed of BCs of 150–555 bp within an eGFP expressing lentiviral vector driven by the Ubiquitin C promoter. A set of 40 BC arrayed lentivectors were produced and the homogeneity of their titers was verify on HEK 293T cells. This BC lentivector library is used to transduce murine HSCs based on the Lin Endo+ Sca.1+ phenotype. Cells are pooled before infusion into myeloablated mice, and the BC signature is analyzed in blood and marrow of transplanted mice at different times after transplantation. After PCR amplification using primers complementary to the neighboring common 5′ and 3′ ends of the BCs, identification of each BC is obtained by gel analysis or by DNA pyrosequencing. Abbreviations: BC, barcode; eGFP, enhanced green fluorescent protein.

Download figure to PowerPoint

This arrayed BC LV library was used to transduce murine Lin- Endo+ Sca.1+ (LES) cells, a pertinent HSC population [[11, 12]]. Cells are transduced in separate wells, with one BC LV per well. When desirable for a given experiment, cells in each well can be exposed to a different condition, such as from an arrayed library of small molecules. Thus, different HSC clones under different conditions can be pooled before their infusion in myeloablated mice. Integrated BCs are replicated and segregated through cell divisions and are present in the HSC progeny in recipient mice. To analyze BCs signature at different time points after transplantation, genomic DNA is extracted from blood or marrow recipient mice, and PCR amplification is performed using primers complementary to the neighboring common 5′ and 3′ ends of the BCs. The PCR efficacy and detection threshold were calculated for amplification of the 150 bp and 555 bp BCs (Supporting Information Fig. S1A, S1B). Results show that the PCR has an efficiency of 105% and 98% for the amplification of the 150 and 555 bp BCs, respectively, thus demonstrating that these BC LVs were suitable for further use of our approach. Furthermore, we are able to detect one single copy of each BC (long and short) by single primer set PCR reactions containing >6,000 diploid mouse genomes, which is the maximum dilution of a BC-bearing genome possible within the 40 ng total genomic DNA used in each of our PCR reactions (Supporting Information Table S1). This result was confirmed by the fact that we were able to retrieve BCs (with appropriate negative control, excluding possible contamination) from several mice in which no or very few donor (CD45.1-eGFP+) cells could be identified by flow cytometry (Supporting Information Fig. S2). The identification of BCs present in blood and BM of transplanted mice, by gel analysis, when only a small number of BC LVs are required for a given experiment, or by DNA pyrosequencing when a large number of BC LVs are needed, enables retrospective analysis of the initial ex vivo HSC input (deconvolution). In the case of agarose gel analysis, BCs are chosen to yield amplicons of different sizes that can be easily resolved by gel electrophoresis.

We also established that this arrayed BC LV library is able to transduce hematopoietic CD34+ cells of humans and non-human primates (Supporting Information Fig. S3). This observation may thus increase the number of preclinical investigations warranted in non-human primates by sparing the number of animals needed, an important issue for reasons of ethics, availability, and cost.

BM HSC Enrichment Based on a Lin Endo+ Sca.1+ Phenotype

A key parameter that underlies the success of the arrayed BC strategy is to isolate relevant cell types and determine the appropriate number of cells that have to be transduced by arrayed BC LVs to yield optimal transplantation efficiencies and deconvolution readout. We chose the LES phenotype to enrich for stem/early progenitor cells because they typify a pertinent hematopoietic population in long-term competitive repopulation assays [[11, 12]]. To determine the functional abilities of these cells and assess their transducibility, we verified that (a) transduction efficiency of LES cells was proportional to MOI, (b) the percentage of donor cells present in recipients after transplantation was proportional to input LES cell dose, and (c) the LES population includes cells with long-term in vivo repopulating potential.

The effect of varying the MOI was determined after transducing LES cells, which represent 0.64% ± 0.12% of the Lin fraction of mouse BM cells (Fig. 2A), with BC LVs. The transduction efficiency was measured by scoring eGFP+ cells by flow cytometry after 7 days of culture. Data demonstrate that increasing the MOI results in an increment of the percentage of eGFP+ cells (Fig. 2B).

image

Figure 2. Characterization of the hematopoietic stem cell enriched LES population. (A): Purification of LES cells: Endo+ Sca.1+ cells sorted from lineage (B220, CD11b, GR1, Ter119, CD8a, CD4, and CD5) negative population. (B): Percentage of eGFP+ LES cells 7 days after transduction at increasing MOI. (C): Percentage of CD45.1 donor-derived cells, in whole blood of primary recipient mice 16 weeks after transplantation and quantified by flow cytometry versus the number of injected cells. (D): Limiting dilution analysis showing the percentage of mice containing fewer than 1% of donor CD45.1 in WBC versus the number of injected cells. Data were obtained by flow cytometry 16 weeks after transplantation. Abbreviations: FSC, Forward Scatter; GFP, green fluorescent protein; LES, Lin Endoglin+ Sca.1+; MOI, multiplicity of infection; SSC, Side Scatter.

Download figure to PowerPoint

We next examined whether increasing the number of LES cells correlates with an increment in the number of donor-derived blood cells in transplanted mice. To this end, LES cells (n = 500–3,000) isolated from CD45.1 mice were transplanted into lethally irradiated CD45.2 mice with 500,000 competitor mononuclear BM cells prepared from CD45.2-eGFP mice. Sixteen weeks after transplantation, blood samples were drawn from transplanted recipients and examined for CD45.1 expression by flow cytometry (Fig. 2C). The percentage of CD45.1 cells found in each transplanted recipient was proportional (R2 = 0.86) to the number of transplanted LES, thereby demonstrating that the LES cell population is suitable for quantifying LT-HSCs in vivo.

The frequency of HSCs in the LES population was determined by a limiting dilution analysis (Fig. 2D). To this end, LES cells (n = 25, 50, 75, and 100) were cultured and then transplanted in myeloablated mice with 500,000 competitor nucleated BM cells. Four months after transplantation, peripheral WBCs were analyzed by flow cytometry. It demonstrated that the frequency of HSC is 1/90 (95% confidence, interval for mean, 1/183-1/45). This was calculated on the basis of Poisson statistics considering as negative the mice containing less than 1% donor-derived cells in total peripheral WBCs.

Quantification of HSC Lineage Biases

To gather further insights into HSC biology, we applied the arrayed BC approach to the topical question of lineage biased hematopoiesis. Four types of murine HSCs have been previously identified on the basis of their differentiation capability: α cells display a high myeloid/lymphoid contribution ratio, β cells show a balanced myeloid/lymphoid contribution ratio, γ cells display a low myeloid/lymphoid contribution ratio, and δ cells contribute B cells and/or T cells but not myeloid cells (<1%) at 16 weeks post-transplantation [13]. To characterize LES cell populations with respect to these differentiation properties, 30 wells containing 50 CD45.1 LES each were individually transduced with a different 150 bp BC (Fig. 3A) and transplanted into myeloablated syngeneic CD45.2 mice.

image

Figure 3. Identification of repopulating cell subtypes α, β, γ, and δ. (A): Single-cell plating; the indicated quantities of CD45.1 LES cells were transduced with 30 different BCs of 150 bp (one BC per well), pooled, and transplanted with eGFP-CD45.2 fresh competitor bone marrow cells into lethally irradiated CD45.2 mice. (B): BC-based scoring of lineage-biased hematopoietic stem cells (HSCs); BC contributions to total peripheral WB as well as My and/or Ly cells are shown individually for the four primary transplanted mice. HSC subtypes are defined by the My/Ly ratios of BC bearing cells; when a BC is only detected in one compartment, My or Ly is mentioned instead of a ratio. The individual BC contributions in transplanted mice are summarized by pies below the tables, with each color representing one BC. (C): BC-derived distribution of α, β, γ, and δ patterns of hematopoiesis 14 weeks post primary transplantation. Abbreviations: BC, barcode; Ly, lymphoid; My, myeloid; WB, white blood.

Download figure to PowerPoint

The BC signature was analyzed by DNA pyrosequencing in total WBCs as well as in myeloid (Gr1+ and Mac1+) and lymphoid (B220+ and CD3+) subpopulations (Fig. 3B). We used a control made of a pool of 30 genomic DNAs derived from cells labeled with only one BC each to normalize the percentage of BC sequences retrieved. Two of the 30 BCs were not detected in the control group and were eliminated from all subsequent analyses (Supporting Information Fig. S1C). Each mouse displayed 6–12 different BCs, with a single of them representing more than 75%. This indicates that, at low HSC input, hematopoiesis is maintained by a small number of relatively dominant clones that coexist with a variety of minority clones. Because we detected a total of 15 different BCs in all transplanted mice, we can state that the frequency of HSCs is at least 1/93, taking the number of plated cells into account. This frequency is similar to values obtained by limiting dilution analysis (Fig. 2D) demonstrating that this arrayed barcoding strategy may substitute for limiting dilution assays. By quantifying α, β, γ, and δ cells using BC ratios in the different subpopulations (Fig. 3B), we obtained frequencies of 23% of α cells, 32% of β cells, 3% of γ cells, and 41% of δ cells (Fig. 3C). This represents 67% of biased cells toward myeloid or lymphoid lineage (Fig. 3C). These results are consistent with single-cell transplantations data [13] but required only four transplanted mice instead of 50.

We then analyzed the fate of HSC subtypes after secondary transplantation (Fig. 4). We transplanted 5 × 106 BM cells from each primary transplanted mouse into one secondary host and verified that barcoded HSCs were able to engraft secondary recipients by flow cytometry analysis of peripheral WBCs of secondary transplanted mice (Fig. 4A). We then analyzed the evolution of each BC pattern between the first and the second transplant 8 weeks after secondary transplantation (Fig. 4B). There was relative stability of BC signatures. The dominant clone was maintained in secondary transplants and coexisted with minor clones previously detected. A loss of complexity was, nevertheless, observed, as more than half of the clones were no longer detectable in secondary recipient mice. Knowing the high sensitivity of our arrayed BC LV approach (Supporting Information Table S1 and Supporting Information Fig. S2), it is likely that this observation reflects a physical reality.

image

Figure 4. Evolution of each barcode (BC) pattern from primary to secondary transplants. (A): Chimerism in primary and secondary recipients 14 weeks and 8 weeks after transplantation, respectively. The percentages of donor-derived barcoded (striped) or nonbarcoded (white) CD45.1 Lin Endoglin+ Sca.1+ cells, of recipient-derived CD45.2 cells (black), of eGFP-CD45.2 competitor cells (gray) were analyzed by flow cytometry. Tx, Transplantation. (B): Time course of BC profiles in primary and secondary mice. Each color represents one BC. (C): Schematic representation of repopulating cell subtypes between primary and secondary transplants. Abbreviation: eGFP, enhanced green fluorescent protein.

Download figure to PowerPoint

Clones of α, β, and δ subtypes generated mostly the same cell types, although their potential was not strictly limited, as they were able to generate other cell types (Fig. 4C). A few undetectable clones in primary recipient mice were able to produce α cells in secondary hosts, thereby demonstrating the recruitment of previously quiescent clones, mainly among myeloid-biased cells. The distribution of the four cell types (α, β, γ, and δ) we have observed is intermediate between that obtained by transplanting freshly isolated cells and cells cultured for 4 days, as follows: the proportion of β and α cells decreases while the proportion of γ and δ cells increases gradually during in vitro culture (Fig. 3C vs. [13]).

Early Hematopoietic Contribution of HSC Clones Is Proportional to Input HSC Dose

After having characterized HSC properties at the single-cell level, we next wanted to analyze hematopoiesis dynamics in competitive mouse transplantation assays by following the differential contributions of each BC in mice transplanted with LES cells transduced ex vivo with a mixture of BC LVs at different input ratios of each BC. In order to make our approach applicable to the deconvolution of HSCs having been exposed to separate ex vivo conditions that may influence HSC behavior before transplantation, we also wanted to define the range of input dose and the time points post-transplantation where linearity is observed.

To determine whether the input ratio between different BCs established in vitro is maintained in the regenerated hematopoietic cells of mice after transplantation, we chose two experimental designs with different levels of complexity using either 4 or 12 different BC LVs (Fig. 5A). To this aim, CD45.1 LES cells were seeded in increasing quantities, transduced with BC LVs, pooled, and then transplanted into myeloablated syngeneic CD45.2 mice with competitive CD45.2 BM. As a transduction control, a duplicate of the experiments was performed to quantify by flow cytometry percentages of BC cells 7 days post-transduction. Results show that the transduction efficiency is stable regardless of the number of LES cells plated (Supporting Information Fig. S4).

image

Figure 5. Hematopoiesis dynamics and relation of BC blood cell output to hematopoietic stem cell input dose. (A): The indicated quantities of CD45.1 Lin Endoglin+ Sca.1+ (LES) cells were transduced with 4 (cell input 1) or 12 (cell input 2) different BCs (one BC per well), pooled, and transplanted with eGFP+ CD45.2 fresh competitor BM cells into lethally irradiated CD45.2 mice. (B): The percentages of donor-derived barcoded (striped) or nonbarcoded (white) CD45.1 LES cells, of recipient-derived CD45.2 cells (black), of eGFP-CD45.2 competitor cells (gray) were analyzed by flow cytometry in WB and BM cells of primary and secondary recipients 8, 16, and 24 weeks post-transplantation. (C): Time course of BC profiles in WB and BM cells of primary and secondary recipient mice. Each color represents one BC. (D): Average of BC percentages associated to each cell input 16 weeks after transplantation. Abbreviations: BC, barcode; BM, bone marrow; eGFP, enhanced green fluorescent protein; WB, white blood.

Download figure to PowerPoint

During a period of approximately 1 year, the BC composition of the hematopoietic population in blood and marrow of primary and secondary recipients was analyzed by flow cytometry (Fig. 5B). The stable percentage of BC-bearing cells over time in primary recipient mice and their presence in secondary recipient mice confirm that HSCs capable of long-term reconstitution were successfully barcoded. To check that the transplantation of barcoded LES cells did not induce any abnormality, the blood count of each mouse was performed (data not shown). We did not detect any lineage bias in the repopulated marrow as the percentages of myeloid and lymphoid cells were the same between barcoded and nonbarcoded blood cell populations (data not shown).

BCs retrieved in transplanted mice were identified and quantified either by PCR resolved by gel electrophoresis, directly or after diagnostic restriction digests for Cell input 1 (Supporting Information Fig. S5) or by DNA pyrosequencing for Cell input 2. To take into account PCR biases caused by GC content and length differences (e.g., 150 vs. 555 bp) among BCs, results were normalized, for each BC, with a corrective factor for differential PCR yield. Each of the experiments contains a control of chromosomally integrated short and long BC mixed in predetermined genomic ratios to derive this corrective factor. Controls were made of the same quantity of genomic DNA derived from cells labeled with only one BC for DNA pyrosequencing of 150 bp amplicons (Supporting Information Fig. S1C). Furthermore, we have shown that the average corrective factor is highly stable over a variety of short/long BC ratios (Supporting Information Fig. S4A and Supporting Information Table S2). It is interesting to mention that our DNA pyrosequencing analysis of PCR products for the deconvolution of a family of 30 BCs of same length (150 bp) containing point mutations showed that biases clearly exist among them with a corrective factor as high as 17 (data not shown), which is of similar order as that measured for short versus long BCs.

The time course of BC signature displayed a robust maintenance of BC representation with only slight variations regardless of the in vitro cell input (Fig. 5C). These results allow us to propose a new model of hematopoiesis (Fig. 6), where hematopoiesis is ensured by clones that are continually recruited and others that are more transient. This best fits a combination of the clonal-succession and continuous-recruitment models of hematopoiesis and excludes a hematopoietic reconstitution governed by stochastic variation only [14].

image

Figure 6. Model of hematopoietic stem cell behavior based on arrayed barcode tagging: a combination of clonal-succession and continuous-recruitment models.

Download figure to PowerPoint

The percentage of BCs associated to each quantity of cell input (500, 1,000, 2,500, and 5,000 LES) was calculated for the two experimental designs. For the one with low levels of complexity, the gel analysis demonstrates that BCs recovered from the regenerated hematopoietic cells keep the proportionality of the cell input dose, suggesting that the method allows for tracking HSCs in a quantitative manner, within the range of LES dose tested and until 4 months post-transplantation (Fig. 5D). In the more complex workflow, gel analysis coupled to DNA pyrosequencing indicates that the method is suitable for tracking HSCs quantitatively, even using more than 10 different conditions (Fig. 5D). As in the first workflow, this is no longer true from 6 months post-transplantation onward because, over time, differences between the BCs associated to different quantities of input cells become smaller and proportionality between in vivo BC recovery and in vitro cell input is no longer linear (data not shown).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

In this study, we developed a HSC labeling strategy based on genomic integration of molecular BCs transferred by an arrayed LV library.

Clonal stem cell tracking to estimate clonal contribution in the hematopoietic system has been performed using single-cell transplantation assays [13], viral insertion sites analysis, especially in the context of human clinical trials [[3, 15]], and non arrayed BC libraries [[7-9]]. Each of these strategies has advantages and limitations (Table 1) and methodological constraints have been discussed in a recent review [14]. Tagging cells with BCs has been previously used to track different cell types simultaneously [[7, 8]], and this technology was recently shown to be useful to replace single-cell analysis of HSC [[9, 10]]. However, the BC libraries used in previous studies were not arrayed. The exact linkage between a given BCs and target cells could not be a priori determined, and complex statistical analysis was required to interpret the data gathered. We show here that an arrayed BC LV library, associated with an easy enriched HSC cell sorting protocol, is a versatile and powerful tool to probe the clonal dynamics of hematopoiesis. In contrast to nonarrayed BC libraries, the actual sequence identity of all the BCs is known in our arrayed library, thereby avoiding the need for estimation of BC complexity and the risk of false identities of retrieved BCs [[7, 9, 16]]. In the future, it may be of value to reduce biases by optimizing BCs of same length with a limited set of point mutations, as in the Hamming Code design approach [16]. This arrayed BC LV approach also avoids certain technical issues that include restriction enzyme biases and random ligand attachment, typically associated with insertion site analysis of chromosomally integrated vectors by linear amplification-mediated- or ligation-mediated-PCR [14].

Table 1. Comparative advantages and limits of existing methods for in vivo clonal stem cells tracking
ApplicationsAdvantagesLimitations
  1. Abbreviation: BC, barcode.

Single-cell assaysQuantitativeHigh number of mice
 Analysis of nonmanipulated cellsNegative mice not tested in second transplant
Viral DNA taggingCompetitive polyclonal analysisPotential bias due to the transduction/culture steps
 QuantitativeSequencing or mapping errors leading to false positive
 Unique genetic markUse of restriction enzymes: lack of accessibility to the whole genome
  Strong PCR bias (length of fragments)
Nonarrayed barcodingLimited number of miceStatistical assumptions required
 QuantitativeNo specific link of a given BC to ex vivo conditions studied
 Competitive polyclonal analysisSecond transplant not performed
 Potential bias due to the transduction/culture steps
Arrayed barcodingLimited number of miceHomogenous individual viral titer required
 QuantitativePotential bias due to the transduction/culture steps
 Competitive polyclonal analysis 
 A priori link of one BC to one ex vivo condition 
 Prospective analysis in second transplant 

We show here that, upon analysis of BCs in the circulating blood of transplanted mice in conditions of limiting dilution, barcoded HSC contribution to hematopoiesis is maintained largely by a small number of relatively dominant clones, although a range of minority clones coexist. Dominant clones found in the primary hosts are also those who are most detected in secondary recipients, demonstrating their strong capacity for self-renewal. These results corroborate those obtained with a nonarrayed BC library [7], by monitoring viral insertion sites [[15, 17, 18]], and in recent human gene therapy clinical trials [3].

Analysis of BCs in myeloid and lymphoid subpopulations confirmed the existence of cells with HSC behavior that are biased toward one of the two lineages. By determining the percentage of each of the BCs in the lymphoid and myeloid lineages of mice engrafted with transduced LES cells, we were able to identify the different cell subtypes previously defined [13]: β cells capable of reconstituting the two lineages, α-cells preferentially yielding myeloid cells, and γ and δ cells preferentially contributing to the lymphoid lineage. Observed percentages of each of these cell types initiated here with LES cells transduced and cultured for 2 days are quite consistent with those obtained by the Eaves group using single-cell transplants [13]. Long-term myeloid-biased hematopoiesis has also been observed in a human gene therapy clinical trial [3] and in other experimental settings [19].

The behavior of these different cell types during serial transplantation was determined by assessing the presence of BCs in secondary mice. More than half of the clones identified in the primary recipient mice are no longer detectable in the secondary recipient mice. This loss of complexity in secondary transplants was also observed by the Eaves group (around 50% of the clones did not engraft) [13] or by the Dick laboratory using analysis of viral insertion sites (57% of clones were only present in primary recipient mice) [20]. Knowing the high sensitivity of our arrayed BC LV approach, it is likely that this observation reflects a physical reality, with the caveat that not all BM cells from primary transplanted mice are injected in secondary transplants.

We also show that the different cell types (α, β, γ, and δ) maintain their fate and properties in secondary transplants, although the potential of β cells is not strictly limited as we observed that these cells are capable of generating γ and δ cells. These results are consistent with previous observations [13] except for the potential of δ cells, which were unable to reconstitute secondary hosts in this previous study [13]. Our own data show here that δ cells, heavily skewed toward the lymphoid lineage, are nevertheless capable of reconstituting a secondary host with a similar profile, thus demonstrating the existence of lymphoid biased HSCs.

We also show here that a few undetectable clones in primary recipient mice are capable of generating a detectable progeny in secondary hosts. This property was also observed by monitoring viral insertion sites in primary and secondary transplants [20]. Although we cannot entirely rule out that this may stem from an insufficient level of sensitivity of clone detection in primary hosts, it suggests the existence of quiescent HSCs that become activated in secondary transplants. This type of clones was not observed in single-cell studies because primary negative mice, which could have been engrafted with a quiescent clone, were not used for secondary transplants [13].

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

This arrayed BC-LV strategy of cell marking combines several advantages provided independently by previously described methods while allowing for a major reduction in the number of mice or non-human primates needed and avoiding complex statistical analysis and interpretation biases. Data obtained here further clarify dynamic models of steady-state hematopoiesis. Our overall results demonstrate that hematopoiesis is ensured by the coexistence of clones that are continuously recruited; a few clones disappear while others are only recruited under particular conditions (e.g., secondary transplants). These results bridge the seemingly opposed clonal-succession and continuous-recruitment models and exclude a model governed by stochastic events [14]. Our data also corroborate the existence of myeloid-lymphoid biases among HSC populations, as reported in mouse transplants [13] and human clinical trials [3]. By further sequencing the BC LV-bearing plasmids available (n = 4,700), a BC LV library of very large complexity can be obtained. Because this is the only method allowing for an a priori association of one BC to one specific ex vivo condition, its direct application to the screening of large libraries of small molecules, with respect to their differential effects on HSCs or tumor cells, is now feasible due to the possibility of in vivo deconvolution and the linearity of the BC representation in vivo in proportion to the input BC dose.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

We thank T. Schumacher (Division of Immunology and Central Microarray Facility, The Netherlands Cancer Institute, Amsterdam) for the E. coli library containing the BCs. We thank A. Cosma and S. Guenounou for cell sorting and S. Prost for helpful discussions. This work was supported by CEA, INSERM, and ANR's Chaire d'Excellence to P.L. J.G. was supported by a fellowship from CEA (Irtélis).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. Disclosure of Potential Conflicts of Interest
  10. References
  11. Supporting Information

Additional Supporting information may be found in the online version of this article.

FilenameFormatSizeDescription
stem1383-sup-0001-suppfig1.tif984KSupporting Information Figure 1
stem1383-sup-0002-suppfig2.tif1512KSupporting Information Figure 2
stem1383-sup-0003-suppfig3.tif485KSupporting Information Figure 3
stem1383-sup-0004-suppfig4.tif449KSupporting Information Figure 4
stem1383-sup-0005-suppfig5.tif2279KSupporting Information Figure 5
stem1383-sup-0006-supptab1.docx11KSupporting Information Table 1
stem1383-sup-0007-supptab2.docx12KSupporting Information Table 2

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.