Rapamycin increases oxidative stress response gene expression in adult stem cells

Magnetic-activated cell sorting enriches undifferentiated male germ cells from rapamycin-treated mice

To examine the effects of mTORC1 inhibition on SSC gene expression, we first implemented an established regimen in which juvenile male mice were administered intraperitoneal injections of rapamycin or control vehicle daily for two weeks (Figure 1A) [4, 12, 24]. Following these treatments, single cell suspensions of germ cells were prepared from isolated testes and subjected to magnetic-activated cell sorting (MACS). This procedure enriches the undifferentiated germ cell fraction, which represents the adult SSC population (Figure 1A) [25-27]. RNA from cells double-positive for the SSC surface markers thymus cell antigen 1, theta (THY1) and glial cell line-derived neurotrophic factor family receptor alpha 1 (GFRA1) was isolated for gene expression microarray analysis and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) validation. In order to verify that our MACS selection strategy was successfully enriching undifferentiated male germ cells, we performed qRT-PCR on MACS-enriched THY1⁺/GFRA1⁺ cells from non-injected mice. When compared to unsorted germ cells, the MACS-enriched cells exhibited a 16-fold increase in Gfra1 transcripts, as well as 10-fold and 8-fold increases, respectively, in two additional SSC transcripts, zinc finger and BTB domain containing 16 (Zbtb16; referred to here as Plzf) and POU domain, class 5, transcription factor 1 (Pou5f1; referred to here as Oct4) (Figure 1B). MACS-enriched THY1⁺ and GFRA1⁺ germ cells have previously been shown to exhibit significantly increased stem cell activity over unsorted cells through transplantation assays [25, 28]. Thus, our experimental design favorably selected for SSCs and allowed for the examination of rapamycin-induced alterations of SSC gene activity.

Rapamycin administration inhibits testis growth but expands the number of undifferentiated germ cells

Testicular weights and body weights of rapamycin-treated mice were both significantly reduced compared to those of control vehicle-treated mice (12.60 ± 0.492 mg vs 33.52 ± 0.968 mg average testis weight; 8.783 ± 0.155 g vs 13.35 ± 0.345 g average body weight; N=5), reflecting the effects of rapamycin on tissue growth (Figure 2 A-D). Administration of rapamycin or control vehicle began on postnatal day (P)12 and ended on P25, a period of mouse development in which the testis generates the first sets of meiotic and post-meiotic germ cells [29]. Equivalent numbers of terminally differentiated Sertoli cells, but fewer mitotic progenitor Leydig cells, were observed between rapamycin- and control vehicle-treated mice (data not shown). Rapamycin had previously been shown to inhibit the proliferation of differentiating germ cells through PI3-K signaling [30]. Conversely, mTORC1 inhibition by rapamycin expands the number of SSCs through glial cell line-derived neurotrophic factor (GDNF) signaling [12]. When we placed SSCs from individual mice into culture, stem cells from the rapamycin-treated mice formed more numerous and larger-sized colonies than the control SSCs during the first two weeks after plating (Figure 3). These results demonstrated that chronic rapamycin exposure selected for adult SSCs within the germ cell population of the juvenile testis. From this, we wondered whether the selection involved transcriptional networks independent of the GDNF signaling pathway.

Identification and validation of a rapamycin-induced SSC transcriptome signature

To identify differentially expressed genes in rapamycin-exposed mouse SSCs, we used the Agilent Whole Mouse Genome oligonucleotide microarray 4x44K platform. Approximately 40% of the SSC transcriptome was significantly altered in expression levels, as measured by >2-fold change and p-value <0.01 (Figure 4A). A total of 7,741 oligo probes were upregulated, while 8,746 probes were downregulated, corresponding to an enhancement of 4,617 transcripts (unique Entrez gene IDs) and a reduction of 5,360 transcripts (Figure 4A, Supplemental Table 1). For data analysis, we used the non-hierarchical clustering method AutoSOME and generated a heat map of the cluster that specifically contained Gfra1 (Figure 4B) [31]. The Gfra1 transcript had previously been shown to be upregulated in SSCs following rapamycin exposure, and it exhibited a 3.13-fold enhancement in expression here (Table 1)[12]. Within the Gfra1 cluster were additional SSC self-renewal-associated genes (Ret, Lin28b,Nanos2, Foxo1), all of which were significantly upregulated in our rapamycin-exposed SSCs (Figure 4B, Table 1). In addition to these genes important for SSC maintenance, the Gfra1 cluster contained several oxidative stress response genes that were also significantly upregulated with rapamycin, including Alad, Sod1, and Gsr (Figure 4B, Table 1). In contrast, genes important for signal transduction in growth and metabolism (Wnt3a, Wnt2, Stat4, Tgfbr1-3) were significantly downregulated in our rapamycin-exposed SSCs (Table 1). To integrate these SSC transcriptome data into biological pathways, we used the knowledge-based database Ingenuity Pathway Analysis^® (IPA). The top five canonical pathways, ranked by p-value and ratio of the number of genes (up/down) per category, are listed in Table 2. These include free radical scavenging, phosphatase and tensin homolog (PTEN) signaling, and nuclear factor (erythroid-derived 2)-related factor 2 (NRF2)-mediated oxidative stress response (Table 2). IPA identified several potential interactions among the oxidative stress response genes in our Gfra1 cluster, with ALAD serving as a nodal point to connect tumor necrosis factor (TNF) and erythroblastic leukemia viral oncogene homolog 2, neuro/glioblastoma derived oncogene homolog (ERBB2) with SOD1, GSR, and retinoblastoma protein (RB1) (Figure 4C). We validated the differential regulation of 15 selected transcripts (9 upregulated: Gfra1, Ret, Lin28b, Nanos2,Foxo1, Alad, Sod1, Gsr, Erbb2; 6 downregulated: Wnt3a, Tgfbr1, Stat4, Tnf, Gsc,Meox2) from the gene expression microarray using qRT-PCR (Figure 5A). All genes exhibited significant changes in expression as predicted from the microarray data.

Oxidative stress response genes upregulated by rapamycin are downregulated in aging SSCs

Given that rapamycin has been shown to increase lifespan in aging mice [13, 19], and that here it significantly enhanced the expression of Alad, Sod1, and Gsr in juvenile SSCs, we next asked whether the levels of these three oxidative stress response transcripts were diminished in the SSCs isolated from older versus younger wild type mice. When SSCs from 55-week-old males were compared to SSCs from 3-week-old males, the relative gene expression values for Alad, Sod1, and Gsr were all decreased (1.46-, 1.72-, and 1.62-fold, respectively; p<0.05) (Figure 5B). Morphologically, the SSCs from the two ages of mice were indistinguishable, although fewer SSCs were obtained from the older testes than from the younger testes (data not shown). These data suggest that as SSCs age in vivo, the aging process correlates with a downregulation in the expression of genes that respond to oxidative stress.

Animals

Male FVB mice aged 12-days-old through 26-days-old were administered daily intraperitoneal (IP) injections of rapamycin (4mg/kg body weight) or control vehicle (5% Tween-80, 5% PEG-400), beginning at postnatal day (P)12. Mice were euthanized at P26 and their testes were isolated for germ cell enrichment. Untreated wild type male FVB mice aged 3-weeks-old and 55-weeks-old were also euthanized for experimental analysis. All procedures and care of animals were carried out according to the Children's Memorial Research Center Animal Care and Use Committee.

Isolation of mouse testicular germ cells

Testes were decapsulated and briefly minced in ice-cold 1:1 Dulbecco's Modified Eagle Medium-Ham's F-12 Medium. An initial enzymatic digestion using collagenase IV (1mg/ml) and DNase I (2mg/ml) at 37°C for 30 min. was administered to remove interstitial Leydig cells and peritubular myoid cells from the seminiferous tubules. A second enzymatic digestion using collagenase IV (1 mg/ml), DNase I (2mg/ml), hyaluronidase (1.5mg/ml), and trypsin (1mg/ml) at 37°C for 30 min. was administered to isolate germ cells and Sertoli cells from the remaining tissue. A final suspension of single cells was prepared in ice-cold PBS containing 0.5% BSA and 2mM EDTA (MACS Buffer) for subsequent spermatogonial stem cell enrichment.

Enrichment of SSCs using MACS separation

The single cell suspension containing germ cells in 80 μl MACS Buffer was first incubated with 20 μl rabbit anti-GFRA1 antibodies (Santa Cruz Biotechnology, CA) at 4°C for 20 min. with rotation. After washes, a second incubation of cells in 80 μl MACS Buffer with 10 μl goat anti-rabbit antibody-conjugated MicroBeads and 10μl anti-THY1 antibody-conjugated MicroBeads (Miltenyi Biotech, Auburn, CA) was administered at 4°C for 20 min. with rotation. The labeled cells were filtered through 30-μm pore size mesh to remove cell aggregates, and then sorted through a separation LS column attached to a MidiMACS separator (Miltenyi Biotec). THY1+ and GFRA1+ cells were retained inside the column within the magnetic field, while unlabeled cells passed through the column and were collected as the THY1-/GFRA1- cell fraction (flow through). After washes with MACS Buffer, the LS column was removed from the magnetic field and the THY1+ and GFRA1+ cells were flushed out.

Establishment of SSC cultures

Approximately 200,000 THY+/GFRA1+ cells were seeded into 35-mm round dishes containing irradiated Mouse Embryonic Fibroblast feeder layers (1.2 x 10⁶ MEFs per dish). Cells were maintained in optimized culture medium [55] (StemPro-34 supplemented with 1% FBS, 10 μg/mL GDNF, 10 ng/mL bFGF, 20 ng/mL EGF, 1,000 units/mL ESGRO/LIF), fed with new media every other day. SSC colonies were visualized using a Leica DM-IRB Inverted Research Microscope.

RNA isolation and qRT-PCR analysis

Total RNA was extracted from MACS-separated cells using the RNeasy Micro Kit (Qiagen, Valencia, CA) following the manufacturer's protocol. RNA samples were treated with RNase-free DNase I (Qiagen) on-column to remove genomic DNA. Yield and quality of RNA samples were determined using the NanoDrop 2000 Spectrophotometer (ThermoScientific, Wilmington, DE). Total RNA was reverse transcribed into cDNA using random hexamer primers (Life Technologies, Grand Island, NY). For qRT-PCR, cDNA was added to 2x Power SYBR^® Green PCR Master Mix (Applied Biosystems, Foster City, CA) with specific oligonucleotide primer sets for the genes of interest (listed in Supplemental Table 2). Samples from three biological replicates were run in triplicate on an Applied Biosystems 7500 Real-Time PCR System using SYBR^® Green dye for read-out and ROX™ dye as an internal reference. Each PCR reaction contained approximately 5-10 ng of cDNA, 1x Power SYBR® Green PCR Master Mix, and 500 nM of each forward and reverse primer for the desired gene. Gapdh was used as an endogenous control. The threshold cycle (C_T), indicating the relative abundance of a particular transcript, was calculated for each reaction by the system software. Quantification of the fold change in gene expression was determined by using the formula 2^−DDCT, in which DDC_T = [(C_T of gene of interest - C_T of Gapdh)_A - (C_T of gene of interest - C_T of Gapdh)_B]. Fold change in transcript levels was plotted using Prism 5 software (GraphPad, La Jolla, CA). Values plotted are mean +/− SEM. Statistical analysis was performed using Prism 5, employing Student's t-test or ANOVA; *p<0.05; **p<0.01; ***p<0.001.

Gene expression microarrays

Total RNA samples were shipped on dry ice to the laboratories at Miltenyi Biotec (Auburn, CA), where they were quality-checked prior to processing with the Agilent 2100 Bioanalyzer platform (Agilent Technologies, Santa Clara, CA). Linear T7-based amplification was performed using 25 ng of each RNA sample. Amplification and Cy3-/Cy5-labeling was performed using the Agilent Low Input Quick Amp Labeling Kit (Agilent Technologies) following the manufacturer's protocol. Yields of cRNA and the dye-incorporation rate were measured using the NanoDrop 2000 Spectrophotometer (ThermoScientific). Hybridization was performed according to the Agilent 60-mer oligo microarray processing protocol using the Agilent Gene Expression Hybridization Kit (Agilent Technologies). The Cy3- and Cy5-labeled cRNAs were combined and hybridized overnight (17 hours at 65°C) onto the Agilent Whole Mouse Genome oligo microarrays 4x44K using a hybridization chamber and oven. Fluorescence signals of the hybridized oligo microarrays were detected using Agilent's DNA microarray scanner (Agilent Technologies). Agilent Feature Extraction (AFE) software was used to process the microarray image files. AFE determines feature intensities and ratios (including background subtraction and normalization), rejects outliers and calculates statistical confidences (p-values). For determining differential gene expression, AFE-derived output data files were further analyzed using the Rosetta Resolver^® gene expression data analysis system (Rosetta Biosoftware, Cambridge, MA). This software was used to generate a double-log scatter plot to visualize the signal intensities of all oligo probes (Figure 3A). AFE-derived output data files contained the gene lists with complete raw data sets.

SSC transcriptome data analysis

The unsupervised clustering method AutoSOME was used to assemble our gene expression data into distinct clusters for subsequent analysis [31]. We used Euclidean distance as a user-defined parameter, scaling our data sets in log2 (establishing a relative value range from 0.0 to 20.0) and employing unit variance normalization, median centering of genes (to eliminate amplitude shifts), and normalizing genes such that the sum of squares of each row/column from our data sets=1. All gene identifiers were collapsed into a non-redundant set by averaging expression values for genes represented by more than one probe. Probes without corresponding gene symbols were not analyzed. The AutoSOME output effectively generated clusters containing gene products assembled according to closely related and interconnected expression patterns. For this study, we focused on the cluster containing Gfra1.

To further analyze the SSC transcriptome, we used Ingenuity Pathways Analysis^® (Ingenuity Systems, Redwood City, CA) to generate networks by uploading our data sets into the application. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base (IPKB). A cutoff value of 2.00 was implemented to identify genes whose expression was significantly differentially regulated, overlaying them onto a global molecular network using information from the IPKB. Networks were algorithmically generated based on their connectivity, generating graphical representations of the molecular relationships among gene products (representing nodes) by depicting them as solid (direct) or dashed (indirect) lines. All lines are supported by at least one reference from the literature or from canonical information stored in the IPKB. Mouse, rat, and human orthologs of genes are stored as separate objects in the IPKB, but are represented as single nodes in the network. Nodes are displayed using various shapes that represent the functional class of the gene product (i.e. parallelogram = enzyme). Canonical pathway analysis utilizes well characterized metabolic and cell signaling pathways that are generated prior to data input and on which identified gene products are overlaid.

The gene expression microarray data in this study are available in the public repository Gene Expression Omnibus; accession # GSE37062.