Human multipotent stromal cells from bone marrow and microRNA: Regulation of differentiation and leukemia inhibitory factor expression

Results

Generation of a Stable Knockdown of Dicer and Drosha in hMSCs.

As a first step toward identifying the role of miRNAs in the differentiation of hMSCs, we analyzed the effect of decreasing the expression of key proteins including the cytoplasmic enzyme Dicer and the nuclear enzyme Drosha. We used a tetracycline inducible vector that contained a short hairpin RNA (shRNA) targeting either Dicer or Drosha to prepare stable cell lines of hMSCs that could be induced to reduce the expression of either enzyme [see supporting information (SI) Fig. S1 a and b]. Transduced cells showed a specific knockdown of Dicer or Drosha at the protein level when induced with doxycycline as compared with transduced hMSCs cultured without doxycycline in the media. Donor matched control cells showed no variation in the expression of these enzymes when cultured in the presence or absence of doxycycline (Fig. S1c).

Dicer or Drosha Knockdown Inhibits Osteogenic Differentiation of hMSCs.

We then tested the effect of impaired miRNA processing and global miRNA knockdown on the ability of hMSCs to differentiate into an osteogenic lineage. The hMSCs transduced with an shRNA targeting Dicer or Drosha were grown to subconfluent levels and then placed in osteogenic media in the presence or absence of doxycycline. Passage matched hMSCs from the same donor, and from the same preparation as the transduced cells, were grown in the same differentiation conditions in the presence or absence of doxycycline as controls. After 28 days, gross and microscopic visualization of transduced hMSCs grown in the presence of doxycycline showed undetectable mineralization, as demonstrated by staining monolayers with alizarin red stain (ARS) (Fig. 1A and Fig. S2a). Both transduced hMSCs grown in the absence of doxycycline as well as untransduced cells grown in the presence and absence of doxycycline showed robust mineralization (Fig. 1A and Fig. S2a). Quantitation of mineralization was performed by measuring absorbance of ARS extracted from stained cultures. Osteogenic monolayers from transduced hMSCs grown in the presence of doxycycline showed a >95% reduction (P < 10⁻⁶) in ARS incorporation compared with transduced cells grown in the absence of doxycycline and control cells grown in the presence and absence of doxycycline (Fig. 1B and Fig. S2b). To further evaluate the effect of Dicer knockdown on osteogenic differentiation of hMSCs, we assayed alkaline phosphatase (ALP) activity in monolayers grown in osteogenic conditions for 15 days. Transduced hMSCs grown in the presence of doxycycline showed a >95% reduction (P < 10⁻⁶) of ALP activity as compared with transduced cells grown in the absence of doxycycline and control cells grown in the presence and absence of doxycycline (Fig. 1C). The level of ALP activity in transduced hMSCs grown in the presence of doxycycline was slightly lower than the level of activity of hMSCs grown in nondifferentiating culture conditions for 15 days.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Dicer knockdown inhibits osteogenic differentiation of hMSCs. hMSCs transduced (T) with a tetracycline inducible shRNA targeting Dicer, as well as control cells (C), were incubated in osteogenic differentiation media in the presence (D) or absence of doxycycline. (A) Representative images of cellular monolayer mineralization on day 28 of osteogenic differentiation. The bottom row is stained with Alizarin Red Stain (ARS), the top row is unstained (magnification 100×). (B) Absorbance of ARS extracts from cellular monolayers on day 28 of osteogenic differentiation. Extracts from nondifferentiated (ND) cells are shown as a negative control. (C) Alkaline phosphatase (ALP) activity of hMSC monolayers. Cellular monolayers were washed and placed in a solution containing p-nitro-phenol phosphate on day 15 of differentiation. Conversion to p-nitro-phenol, a substance that absorbs light at 410 nm, was measured. For B and C, the average of six culture replicates with Standard Deviation (SD) is shown (raw average and standard deviation numbers are located in Table S2). (D) Relative level of osteopontin and osteocalcin mRNA in cells differentiated for 21 days. Values are normalized to GAPDH mRNA levels and subsequently to mRNA levels of undifferentiated cells (U). The average of three replicates is displayed.

We then assessed the level of osteopontin and osteocalcin, two late markers of osteogenesis, in cells that had been differentiated for 21 days as compared with undifferentiated cells. hMSCs transduced with an shRNA targeting Dicer differentiated in the presence of doxycycline had a small increase in osteopontin and osteocalcin (3.95- and 2.8-fold, respectively), when compared with undifferentiated cells (Fig. 1D). In contrast, transduced hMSCs grown in the absence of doxycycline had significantly greater changes in mRNA levels (P < 0.0001) with a 34.3-fold increase in osteopontin and 14.8-fold increase in osteocalcin, as compared with undifferentiated cells.

Dicer or Drosha Knockdown Inhibits Adipogenic Differentiation of hMSCs.

A similar approach was used to assess the ability of hMSCs to differentiate into an adipogenic lineage after inhibition of Dicer or Drosha. Transduced hMSCs were grown to subconfluent levels and then placed in adipogenic media in the presence or absence of doxycycline along with passage matched controls. After 21 days, bright field microscopic inspection of oil red-o (ORO) stained cultures demonstrated that hMSCs transduced with either an shRNA targeting Dicer or Drosha grown in the presence or absence of doxycycline and control hMSCs grown in the presence and absence of doxycycline were able to form intracellular lipid droplets (Fig. 2A and Fig. S3a). The intracellular lipid accumulation in transduced hMSCs grown in the presence of doxycycline was sparse, often containing small droplets that did not coalesce, a characteristic of early lipid formation. Transduced hMSCs grown in the absence of doxycycline and control hMSCs grown in the presence and absence of doxycycline showed accumulation of larger lipid droplets, characteristic of later stage adipogenesis. Quantitation of lipid formation in the Dicer knockdown model, by measuring absorbance of ORO extracted from stained cultures, showed a 70% reduction (P < 0.001) in transduced hMSCs grown in the presence of doxycycline when compared with transduced hMSCs grown in the absence of doxycycline and control cells grown in the presence and absence of doxycycline (Fig. 2B). In the Drosha knockdown model, a >95% reduction of lipid droplet formation (P < 10⁻⁶) was observed (Fig. S3b). To further evaluate the effect of Dicer knockdown on adipogenic differentiation of hMSCs, we evaluated hormone sensitive lipase (HSL) activity in monolayers grown in adipogenic conditions for 21 days using a technique first described by Wieland (12). Both transduced hMSCs grown in the absence of doxycycline and control hMSCs grown in the presence of doxycycline showed lipolytic activity that increased when increasing concentrations of isoproterenol were added to the cellular monolayers (Fig. 2C). In contrast, the transduced hMSCs with an shRNA targeting Dicer grown in the presence of doxycycline showed significantly less baseline lipolytic activity (P = 0.03). Furthermore, lipolytic activity was unaffected by increased isoproterenol concentration, with significantly lower lipolytic activity in Dicer knockdown hMSCs after stimulation with 5 nanomoles of isoproterenol (P < 0.0001) (Fig. 2C). Thus, despite the ability of hMSCs to form some lipid droplets with impaired miRNA processing, the hormone sensitive lipolytic functional activity characteristic of adipocytes remained very low if not absent, suggesting adipogenic differentiation was either disrupted or severely retarded.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Dicer knockdown inhibits adipogenic differentiation of hMSCs. hMSCs transduced (T) with a tetracycline inducible shRNA targeting Dicer, as well as control cells, were incubated in adipogenic differentiation media in the presence (D) or absence of doxycycline. (A) Representative images of lipid droplet formation on day 21 of adipogenic differentiation. The bottom row is stained with Oil Red-O (ORO) and the top row is unstained (magnification 100×). (B) Absorbance of ORO extracts from cellular monolayers on day 21 of adipogenic differentiation. The average of six culture replicates with SD is shown. (C) Hormone sensitive lipase (HSL) activity of hMSC cellular monolayers cultured in adipogenic media for 21 days. Monolayers were washed and incubated at varying concentrations of isoproteranol for 2 h. Glycerol released into the medium was measured by assessing absorbance at 350 nm, based on conversion of NAD to NADH. The average of three culture replicates with SD is shown. For B and C, raw average and standard deviation numbers are located in Table S2 (D) Relative level of FABP4 and adiponectin mRNA in cells differentiated for 21 days. Values are normalized to GAPDH mRNA levels and subsequently to mRNA levels of undifferentiated cells (U). The average of three replicates is displayed.

We then assessed the level of adiponectin and FABP4 transcripts, two late markers of adipogenesis, in cells that had been differentiated for 21 days as compared with undifferentiated cells. hMSCs transduced with an shRNA targeting Dicer differentiated in the presence of doxycycline had a 5.6-fold increase in adiponectin with a 50% decrease in FABP4 when compared with undifferentiated cells (Fig. 2D). In contrast, transduced hMSCs differentiated in the absence of doxycycline had a significantly larger increase in both mRNA levels (P < 0.0001) with a 250-fold increase in adiponectin and 6.4-fold increase in FABP4 compared with undifferentiated cells.

miRNA Expression During Differentiation of hMSCs.

Next, we performed miRNA expression analysis of hMSCs from three preparations from each of three donors as they differentiated into both adipogenic and osteogenic lineages. We used a new bead based flowcytometric method to analyze 435 known miRNAs, demonstrated by Lu and colleagues to be specific to a 1 nucleotide mutation (13). Of the 435 miRNAs tested, 58 were expressed in all three donor cell populations in undifferentiated hMSCs (Table S1). Stringent expression analysis to exclude changes reflecting proliferation rate and cell cycle status showed 19 miRNAs were up-regulated during osteogenic differentiation (Fig. 3A Right) and 20 were up-regulated during adipogenic differentiation (Fig. 3A Left). Of the miRNAs that were up-regulated during differentiation, 11 were up-regulated under both conditions (Fig. S4). Only one miRNA decreased in expression during adipogenic differentiation.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

miRNA expression during hMSC differentiation. (A) Heat map of miRNA expression during differentiation toward osteogenic and adipogenic lineages (Red, high expression; Blue, low expression). Three different donors were cultured in specific differentiation media and RNA was collected at Day (D) 0 (no differentiation), 1, 3, 7, and 14. Only miRNAs showing differential expression between early time points (days 0, 1, and 3) and late time points (days 7 and 14) are shown (90% CI P < 0.01). (B) Table identifying five miRNAs that are up-regulated during osteogenic or adipogenic differentiation and are potential regulators of leukemia inhibitory factor (LIF). Predictions from three computer-based algorithms are shown. (C) Real-time RT-PCR analysis of six miRNAs up-regulated in adipogenic differentiation (A) and osteogenic differentiation (O). Relative miRNA levels of Dicer knockdown (KD) cells are also shown. Values are normalized to 18S rRNA expression and subsequently to miRNA levels of undifferentiated cells (U). The average of four replicates is displayed.

We then searched for mRNA targets that were known to be important in regulating cellular differentiation and that may be regulated by the candidate miRNAs. Using three in silico prediction algorithms (14–16) designed to identify miRNAs that target the 3′ UTR of mature transcripts, we discovered that five of the miRNAs that were up-regulated in hMSCs during differentiation were predicted to target LIF: hsa-mir 26a, hsa-mir 125a, hsa-mir 125b, hsa-mir 199a, and hsa-mir 346 (Fig. 3B). When hMSCs were cultured in media was supplemented with exogenous LIF, we observed a 46% reduction (P < 0.05) in lipid accumulation on day 21 of adipogenic differentiation, compared with adipogenic cultures without supplemental LIF (Fig. S5 a and b). We also observed a 43% reduction (P < 0.05) of ALP activity on Day 7 of osteogenic cultures containing supplemental LIF, compared cultures without supplemental LIF (Fig. S5c).

To confirm that the five candidate miRNAs that targeted LIF were up-regulated during differentiation of hMSCs, we performed real time RT-PCR by using a stem loop priming method (17). All five miRNAs were up-regulated at day 15 of differentiation in both osteogenic and adipogenic conditions when compared with a subconfluent undifferentiated sample (P < 0.001) (Fig. 3C). In addition a sixth miRNA, hsa-mir23a, that has been demonstrated to regulate stromal derived factor 1 (SDF1) (18) was up-regulated during both osteogenic and adipogenic conditions (Fig. 3C). Furthermore, cells deficient in Dicer also showed a significantly lower level of all 6 miRNAs when compared with the undifferentiated sample of hMSCs (P < 0.001), indicating that the cells deficient in Dicer also showed a decrease in miRNA processing (Fig. 3C).

miRNAs Target LIF During Differentiation of hMSCs.

Next, we examined functional modulation of LIF in differentiating hMSCs. First, we analyzed the level of LIF secreted in culture media as hMSCs differentiated, which correlates with increase in miRNA expression. As hMSCs differentiated, the level of secreted LIF in the culture media decreased (Fig. 4A). Furthermore, hMSCs transduced with an shRNA targeting Dicer in the presence of doxycycline showed significantly higher level of LIF expression during differentiation than transduced cells incubated in the absence of doxycycline. These data demonstrated that the LIF decrease during differentiation is coincident with the increase in miRNAs predicted to target LIF; and that when the miRNA processing mechanism is disrupted these changes in LIF secretion are attenuated.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

miRNA regulation of LIF in hMSCs. (A) ELISAs of LIF in hMSC conditioned media during osteogenic and adipogenic differentiation. hMSCs were incubated for 48 h to condition media during the specified time points. The average of four culture replicates is displayed with SEM. (B) ELISAs of LIF in conditioned media of hMSCs following transfection with specific synthetic miRNAs, scrambled siRNA (SCR), all five miRNAs (all), mir199a and mir346 together or a sham procedure (SHAM). After the transfection procedure hMSCs were incubated for 48 h to condition the media. The average of 8 culture replicates is displayed with SD. (C) Immunocytochemistry of hMSCs 36 h after transfection with synthetic miRNAs and controls, as noted above. Green indicates LIF. Cells were counterstained with DAPI (blue) and Alexafluor 594 (red) conjugated phalloidin to identify nuclei and actin cytoskeleton, respectively. (D) Luciferase activity of A549 cells cotransfected with a luciferase reporter plasmid containing the 3′UTR of LIF along with synthetic miRNAs, scrambled siRNA or nothing. The average of four culture replicates is displayed with SD. For B and D, raw average and standard deviation numbers are located in Table S2.

To prove a direct role of specific miRNAs in the modulation of LIF expression, we transfected hMSCs with synthetic forms of all five miRNAs predicted to target LIF, hsa-mir199a hsa-mir346, hsa-mir26a, hsa-mir125a, and hsa-mir125b. Following transfection, the secretion of LIF was significantly decreased in cells transfected with miRNA hsa-mir199a and hsa-mir346 (Fig. 4B) (p <. 0001). Cells transfected with hsa-mir26a, hsa-mir125a, and hsa-mir125b showed no significant alteration in LIF secretion. When hsa-mir199a and hsa-mir346 were cotransfected at half the concentration into the same hMSC population, a greater reduction in LIF secretion occurred than with either synthetic miRNA alone (P < 0.001) (Fig. 4B). The reduction of LIF secretion was confirmed by performing immunofluorescent analysis of cellular LIF levels in cells transfected with both hsa-mir199a and hsa-mir346 (Fig. 4C). Next, we tested the role of hsa-mir199a and hsa-mir346 in targeting the 3′UTR of LIF using a luciferase reporter gene with LIF 3′UTR sequences. Cotransfection of hsa-mir199a or hsa-mir346 and a luciferase reporter construct containing the 3′ UTR of LIF, resulted in a >80% reduction of luciferase activity (P < 0.001), compared with control cells (Fig. 4D).