Pasteurization of human milk affects the miRNA cargo of EVs decreasing its immunomodulatory activity

In this report, we evaluated the effect of the pasteurization (P) process of mother’s own milk (MOM) on the miRNA content of extracellular vesicles (EVs) and its impact on innate immune responses. Differences in size or particle number were not observed upon pasteurization of MOM (PMOM). However, significant differences were observed in the EV membrane marker CD63 and miRNA profiles. miRNA sequencing identified 33 differentially enriched miRNAs between MOMEV and PMOMEV. These changes correlated with significant decreases in the ability of PMOMEV to modulate IL-8 secretion in intestinal Caco2 cells where only MOMEV were able to decrease IL-8 secretion in presence of TNFα. While EVs from MOMEV and PMOMEV were both able to induce a tolerogenic M2-like phenotype in THP-1 macrophages, a significant decrease in the transcript levels of IL-10 and RNA sensing genes was observed with PMOMEV. Together, our data indicates that pasteurization of MOM impacts the integrity and functionality of MOMEV, decreasing its EVs-mediated immunomodulatory activity. This data provides biomarkers that may be utilized during the optimization of milk processing to preserve its bioactivity.


Results
Pasteurization does not affect the EVs yield of MOM. To evaluate the role of pasteurization on the stability of miRNAs in human milk, we first developed PMOM (as a proxy for DHM) in which the components are identical to MOM, except that it has been subjected to pasteurization (see "Material and methods" for details). Next, the EVs enriched fractions of MOM and PMOM were assessed for yield and size distribution. A comparable yield of EVs was observed between MOM EV and PMOM EV with an average concentration of 1.6 × 10 10 ± 9.6 × 10 8 and 1.7 × 10 10 ± 1 × 10 9 EVs per mL of milk, respectively (Suppl. Fig. 1). MOM EV and PMOM EV also showed a similar size distribution (200 nm in average diameter), and negative zeta potential (MOM EV = − 13.4 mV, PMOM EV = − 16.64 mV, respectively). However, immunoblots using the exosome common markers CD63 and CD9 showed that only CD9 was detected in MOM EV and PMOM EV. On the contrary, CD63 was observed in MOM EV samples but was absent in PMOM EV , indicating damage or degradation during the pasteurization process. The ER marker Calnexin was not detected in either EVs samples, however the apolipoprotein A1 was detectable in both MOM EV and PMOM EV , indicating that lipoproteins were co-purified with human milk EVs (Fig. 1).
MOM EV and PMOM EV are enriched with different miRNA cargo. Global miRNA RNAseq was performed on MOM EV and PMOM EV enriched samples. A total of 166 miRNA were identified with 33 miR-NAs of those being differentially enriched. It was found that 9 miRNAs were enriched in PMOM EV while 24 were enriched in MOM EV (Fig. 2, Suppl. Table 1). To further understand the potential impact of the differences observed between MOM EV and PMOM EV , we utilized the microRNA target filter tool using the Ingenuity Pathway Analysis software (IPA) ( Table 1). Of the differentially enriched mRNAs, 25 were predicted to target 201 messenger RNAs. About fifty percent of those genes (92 mRNAs), are involved in five different pathways: AMPK signaling, NF-κB signaling, STAT3 pathway, T Cell Receptor signaling, and IL-15 production (Fig. 3). The analyses of those five pathways showed that they share 28 genes that may be affected by the miRNAs differentially enriched in MOM EV and PMOM EV .
Specifically, the miRNAs enriched in MOM EV target 41 mRNA while the miRNAs enriched in PMOM EV target 51 genes (Suppl. Table 2). To integrate and visualize the IPA results obtained, we used Circos plot to combine in a single plot the potential pathways affected by miRNA from both MOM EV and PMOM EV (Fig. 4). As shown in the Circos plot, only has-let-7b-5p, has-miR-16-5p, has-miR-193a-5p, and has-miR-423-5p identified in PMOM EV are predicted to have an impact on these pathways while 14 miRNAs from MOM EV are predicted to impact the mRNA involved in the same pathways (Figs. 3 and 4, Table 1 and Suppl. Table 2). Based on these predictions, miRNAs enriched in PMOM EV may have a greater influence in the regulation of the transcription factor NF-κB   (Fig. 4). However, both pathways are essential in the development and modulation of multiple aspects of the innate and adaptive immune system 22,23 . These results suggest that miRNA differentially enriched in MOM EV and PMOM EV may affect the immune system through different but overlapping mechanisms.

EVs from MOM and PMOM stimulate IL-8 secretion in Caco2 cells.
Based on the previous predictions, the impact of enriched MOM EV and PMOM EV fractions was evaluated on the nuclear factor kappa B (NF-κB) pathway by following the induction of IL-8 expression 24 . As controls, whole MOM and PMOM as well as exosome free supernatants were tested. A total of 24 fresh MOM samples were pooled into 3 groups and their ability to stimulate IL-8 secretion in Caco2 cells was evaluated. Differential fractionation was performed to evaluate whole milk, EV free supernatant and EV enriched fractions. It was found that while all the treatments induced secretion of IL-8 in Caco2 cells compared to the control, significantly higher secretion levels of IL-8 were observed in Caco2 cells treated with MOM 2%, supernatant and MOM EV when compared to their corresponding PMOM fractions (Fig. 5A). The results obtained with the supernatants were expected as the milk contains a variety of cytokines that may nonspecifically stimulate IL8 secretion. The effect of the human milk EVs was further investigated in a proinflammatory environment by adding TNFα to Caco2 cells in presence or absence of EVs. It was found that MOM EV had similar secretion levels of IL-8 as the TNFα control, while the addition of TNFα to Caco2 in presence of MOM EV was able to decrease IL-8 secretion. In contrast, PMOM EV showed trends (albeit not significant p value = 0.055) to increase IL-8 in presence of TNFα in Caco2 cells, (Fig. 5B). TNFα is a positive regulator of IL-8 gene expression through NF-κB, which is essential for IL-8 gene transcription 24 . These results are in agreement with predicted regulatory effect of PMOM EV miRNAs on the NF-κB signaling pathway (Fig. 4).
Another predicted signaling pathway affected by the miRNAs enriched in both MOM EV and PMOM EV is STAT3 Pathway (Figs. 3 and 4). The Signal Transducer and Activator of Transcription (STAT) 3 pathway is involved in survival, cell growth, and immune response 25 . This pathway is rigorously regulated by Janus Kinase (JAK) and Epidermal Growth Factor Receptor (EGFR). We identified EGFR as one of the targets regulated by www.nature.com/scientificreports/ hsa-miR-16-5p enriched in PMOM EV . In order to elucidate and further understand if the EVs from human milk were able to differentially modulate these pathways, we determined the mRNA expression of some key effectors such as AKT, STAT3, and IGF-1 receptor, in Caco2 cells treated with MOM EV or PMOM EV . A significant decrease in the expression of IGF-1 receptor was observed with both EVs treatments while no significant differences were observed in the expression of AKT compared to the control ( Fig. 5C-D). Additionally, a 3.5-fold increase was observed in the expression of STAT3 after treating Caco2 cells with MOM EV . While the expression of STAT3 was higher in the cells treated with PMOM EV, no significant differences were observed when compared to the control (Fig. 5E). However, when STAT3 was evaluated by Western blot, no significant differences were observed between the treatments and the control ( Supplementary Fig. 2). These results suggest that regulation of STAT3 transcript could be modulated in a hsa-miR-16-5p through EGFR repression in Caco2 cells treated with PMOM EV .
Differential M2 tolerogenic differentiation of THP-1 human macrophages by MOM EV and PMOM EV . The potential impact of the enrichment of has-let-7b-5p in PMOM EV was evaluated on the mRNA expression levels of IL10 26 using the THP-1 PMA-activated macrophage cell line. Macrophages are broadly located in the human body playing a crucial role in regulating immune responses 27 . Once activated, macrophages can differentiate into two sub-types M1-like macrophages capable of proinflammatory responses and M2-like macrophages capable of tolerogenic responses 28 . To this end, THP-1 monocytes induced with PMA were treated with MOM EV and PMOM EV at 10 10 particles/well for 6 h. The mRNA levels of IL-10, TNFα and IL-1β were quantified as signature markers of tolerogenic THP-1 (M2) macrophages. It was found that both MOM EV and PMOM EV induced a M2 phenotype when compared to the vehicle control. Nonetheless, THP-1 macrophages  www.nature.com/scientificreports/ treated with MOM EV had significantly higher levels of IL-10 expression when compared to PMOM EV ( Fig. 6A-C). These findings are in agreement with the higher abundance of has-let-7b-5p in PMOM EV . As lipoproteins were observed with MOM EV and PMOM EV , their contribution to the M2b phenotype observed was investigated. LDL (Low-density lipoprotein) and HDL (High-density lipoprotein) lipoproteins from MOM EV and PMOM EV enriched fractions were extracted and tested at equivalent concentrations to those found in the EVs. The addition of MOM HDL or PMOM HDL significantly induced expression of TNF-α, IL1β and IL-10 when compared to the vehicle control ( Fig. 6A-C). However, the induction levels were significantly lower than those compared to the MOM EV or PMOM EV . It was found that the addition of MOM LDL or PMOM LDL resulted in a similar response as the vehicle control for all the genes tested. These results indicate that lipoproteins found in the EV enriched fractions partially contribute to the tolerogenic M2b phenotype.
Next, we investigated the impact of milk EVs on the IL-15 production pathway (Figs. 3 and 4). IL-15 is a master regulator playing a crucial role in inflammatory and protective immune response 29,30 . IL-15 has pivotal role as a signal molecule shared by several pathways like JAK/STAT pathway, Ras/MEK/MAPK pathway, PI3K/ Akt/mTOR pathway and NF-κB pathway 31 . Interestingly, significant expression levels of IL-15 mRNA were observed in THP-1 macrophages treated with MOM EV but not with PMOM EV (Fig. 7 A) or with purified lipoproteins (Suppl. Fig. 4). These results are in agreement with the predicted regulatory role of has-let-7b-5p and has-miR-16-5p, enriched in PMOM EV, on the NF-κB signaling pathway.
IL-15 in humans is further regulated by IFN-γ (also known as IL-29) which is part of the IL-10-family of cytokines. This type III interferon exhibits similar type I interferon-like RNA sensing properties resulting in activation of the JAK/STAT pathway 32,33 . Additional inducers of interferon are the sensing of RNA by retinoic acid-inducible gene I (RIG-I) 34 and 2'-5'-oligoadenylate synthetase 2 (OAS2). Therefore, to further evaluate the impact of differential miRNA cargo in human milk EVs, in RNA sensing pathways and interferons, the mRNA www.nature.com/scientificreports/ levels of IL29, OAS2, RIG-I and INF-α were determined. Although both treatments, MOM EV and PMOM EV , induced higher expression of these genes when compared to the control, significant higher levels of OAS2, INF-α and IL29 were observed in THP-1 macrophages treated with MOM EV compared to PMOM EV for p < 0.05 (Fig. 7). Altogether, these results indicate that the miRNA cargo of human milk EVs could be compromised after the pasteurization process decreasing or changing the regulatory effect exerted by human milk miRNA.

MOM EV differentially affects AMP, IKBα and STAT3 signaling pathways. The impact of MOM EV
and PMOM EV on AMPK signaling, NF-κB signaling (by following the protein basal level and activation of the IkB kinase), and STAT3 pathways, was confirmed by evaluating the protein levels in THP-1 macrophages. It was found that the basal level abundancy of ERK and IkBα were significantly higher in MOM EV when compared to the control (Fig. 8). STAT3 showed a similar trend albeit not statistically significant. PMOM EV showed significantly lower concentrations of STAT3 and ERK when compared to MOM EV while IkBα showed a similar trend albeit not statistically significant. The quantification of the activating phosphorylation's in ERK (P-Thr202/ Tyr204 ERK/ERK) showed a significant decrease in PMOM EV when compared to MOM EV . No significant differences were observed for P-S727 STAT3/STAT3 between MOM EV, PMOM EV and the vehicle control (Fig. 8).
These results are in agreement with the stronger stimulation of innate immune responses observed with MOM EV .

Discussion
Many mothers that deliver preterm do not produce sufficient volumes of MOM 35 , and as a result, DHM is provided for these infants due to the increased benefits over formula 6 . However, the pasteurization process of DHM significantly reduces the availability of bioactive molecules as well as personalized live commensal bacteria 9 . In this report, we evaluated the impact of pasteurization on human milk on EVs integrity and miRNA cargo. We found that while similar yields and charge of EVs were obtained for MOM and PMOM EVs, a significant decrease in the protein marker CD63 was observed in PMOM EV , while CD9 was detectable in both EVs samples. Other  36,37 . The presence of lipoproteins could explain the comparable yield and size between MOM EV and PMOM EV samples. Recent reports in bovine milk-derived EVs evaluated the impact of several industrial processes on EV integrity. Similar to our findings, pasteurization and homogenization significantly affected the levels of CD63 while CD9 was heat stable 38 . Other studies have shown that the stability of exosomes is highly affected by low temperatures as well 39 . www.nature.com/scientificreports/ Our RNAseq approach also identified significant changes to the miRNA diversity in EVs of MOM upon pasteurization. These results suggest that the decrease in protein stability of the EVs was translated into changes in the miRNA cargo in human milk. The stability and shifts in the miRNA cargo upon heat treatments procedures such as pasteurization and ultra-heat treated (UHT) milk has been studied in human and bovine milk with conflicting results 38 . Smyczynska et al. 18 observed more severe effects of pasteurization of DHM, using Holder Pasteurization (HoP) than with High-Pressure Processing (HPP). A 302-fold decrease in the yield of exosomes was observed with (HoP), with a complete loss of RNA fragments of length typical for miRNA and piRNA. High pressure processing DHM showed less detrimental effect than pasteurization in human milk 18 . However, the impact of these results on host responses were not evaluated. Some studies have reported a significant effect in the abundance and stability of miRNA in bovine milk 38,40 , while others have shown that pasteurization does not have a significant effect 41,42 . These conflicting results may be explained by the differences in methods used for their analyses.
We determined the differential enrichment of 33 miRNA between MOM EV and PMOM EV . Moreover, the main five pathways predicted to be impacted by the differentially enriched miRNA share multiple genes resulting in overlapping effects on immune stimulation and cell proliferation. To evaluate the impact of the potential loss in integrity or decreased abundancy of specific miRNA, we used intestinal epithelial and macrophage cell lines to measure key gene markers. The miRNA enriched in MOM EV and PMOM EV target 25 genes involved in NF-κB signaling pathway. Of those, only six genes are shared among MOM EV and PMOM EV (CD40, CXCR5, PIK3R3, TGFBR1, TGFBR2 and TLR4). We hypothesized that the differential enrichment of miRNA after pasteurization would affect signaling through the NF-κB pathway. To this end, we followed the expression of IL-8 in human epithelial cells 24 . Stimulation of IL-8 secretion was observed with all the fractions of human milk tested (human milk, supernatant free of EVs and EVs), however, significant difference in the stimulation of IL-8 secretion was observed when fractions from PMOM were compared with their correspondent MOM fractions. Similar findings have been observed in the stimulation of IL-8 secretion after heat treatment of bovine and goat milk EVs 43,44 . Likewise, a reduction in the expression of IL-6 was observed in a mice model of necrotizing enterocolitis (NEC) when comparing raw and pasteurized human milk EVs 45 .
Another predicted pathway showing overlapping effector genes was the MAPKs and PI3K/AKT. To further investigate the effects of MOM EV and PMOM EV in Caco2 cells, we evaluated the expression levels of 3 key genes involved in these pathways. Lower expression levels of IGF1R were observed with both MOM EV and PMOM EV treatment in Caco2 cells. However, only MOM EV showed statistically significant differences. IGF1R activation mediates signaling cascades through MAPKs and PI3K/AKT 46,47 impacting many cellular responses including cell proliferation, differentiation, and survival 46,47 . We found that the posttranscriptional regulators of IGF1R, hsa-miR-16-5p and hsa-let-7b-5p, were enriched in PMOM EV , consistently with the lower levels in IGF1R mRNA observed. However, the decrease in expression did not reach statistical significance. The indirect compensatory effect of increased expression levels of STAT3 observed could explain those observations. www.nature.com/scientificreports/ The beneficial effects of human milk on immune stimulation have been largely described 1,2 , however little is known about the immune stimulatory properties of human milk EVs. In bovine milk, the lactation-related differential expression of miRNAs suggests that the miRNA produced in the mammary glands may have a specific function 48 . Since many of the enriched miRNA in human milk EVs identified in this work are predicted to impact immune modulatory pathways, we evaluated the impact of MOM EV and PMOM EV on the stimulation of a M2-like tolerogenic phenotype in macrophages. Interestingly, a stronger stimulation of IL-10 (as gene expression levels) was observed in macrophages treated with MOM EV compared to PMOM EV . These results concur with the predicted activation of the canonical NF-κB signaling pathway through has-let-7b-5p and has-miR-16-5p enrichment quantified in PMOM EV . While the tolerogenic M2-like polarization of macrophages observed in this work has been described earlier for other miRNAs 17,49 , macrophage differentiation and responses could be tissue-and species-dependent. In example, inflammatory M1-like macrophage differentiation, characterized by high levels of IL-6, TNFα, IL-12/23 and decreased IL-10, was observed with bovine milk-derived EVs (BEVs) using a mice model for agricultural dust exposure 50 . In contrast, BEVs administration in two mouse models for arthritis reduced serum levels of MCP-1 and IL-6 correlated with delays in the onset of arthritis and diminished cartilage pathology and bone marrow inflammation 51 .
A remarkable finding was the induction of IL15 in macrophages treated with MOM EV . This cytokine has protective roles towards viral and bacterial infections 52,53 . IL15 has been also successfully used as adjuvant in many antiviral vaccines 23 . The stimulation of IL15 positively correlated with higher expression levels of the interferons IL29, INFα, and the RNA sensing genes RIG1 and OSA2 in MOM EV when compared to PMOM EV . The potential role of milk miRNA in viral interference was explored through in silico methods in the context of the SARS-CoV-2 pandemic. It was found that some of the abundant miRNA in milk (miR29a, miR21 and miR181), can interfere with replication of a wide variety of viruses such as HIV, enterovirus 71 and influenza. However, scarce information is available on the mechanisms that mediate these cross-kingdom interactions 11,19 . Together, these results suggest that human milk EVs may play a significant role in stimulation of RNA sensing pathways to trigger an antiviral response that may be diminished by the pasteurization process. www.nature.com/scientificreports/ There is substantial evidence that milk exosomes can be absorbed and delivered to peripheral tissues where they can exert their regulatory effect (For review see Zempleni et al.) 12 . Therefore, the differences observed in the bioactivity of MOM after pasteurization raise concerns regarding the potential impacts of these processes on health outcomes. While the benefits of MOM have been widely reported, the benefits provided by specific bioactive components such as EVs and their cargo in human milk is unknown. It has been shown that physiological concentrations of bovine milk miRNA can affect gene expression in vivo and in cell cultures (Peripheral Blood Mononuclear Cells, HEK-293 Kidney) 54 . These reports suggest that a decrease in MOM EVs concentration or in their integrity affecting their cargo bioactivity can potentially have an impact on the health of infants. The most significant difference observed in this work was a decrease in the stimulation of IL15 as well as RNA sensing genes by PMOM EV . The reduced stimulation observed in innate immune responses maybe translated into a decreased response to viral infections. These findings, combined with the reported decreases in numerous www.nature.com/scientificreports/ proteins (lactoferrin, lysozyme) 55 , immunoglobulin, and cytokines (like IL-7) 56 reported by others 9,57 , may significantly impact the overall immune bust usually provided by MOM.
The results presented here highlight the need to optimize processes in human milk banks in order to preserve the potential bioactivity of all components in human milk while maximizing its biosafety. There is no direct substitute for the nutritional and immune benefits that MOM provides, but its proven benefits highlight the need for attempting to replicate these factors in alternative feedings as closely as possible. In this work, we identified mechanistic effectors that may be utilized as biomarkers for process optimization. The decreased immunomodulatory activity of pasteurized milk and its derived EVs observed needs to be addressed in future studies, in order to establish better processing strategies and to provide our VLBW infants with the best and personalized nutrition possible.

Material and methods
Milk collection and EV enrichment. For the RNAseq extractions, samples were collected using a sterile Symphony® double breast pump kit at a single expression session with an electric hospital Symphony® breast pump (Medela, McHenry, IL). The protocols used in this study for sample collection were reviewed and approved through the University of Florida Institutional Review Board (IRB RB201400527), a written informed consent was obtained from each donor mother. All methods here were performed in accordance with the relevant guidelines and regulations of the University of Florida Institutional Review Board. Beyond standard hand washing and pumping per NICU protocol, no breast hygiene preparation was performed. The sample was assigned a de-identified subject number, then immediately transported on ice to the microbiology lab for further processing. For the functional analyses using human cell lines, 24 milk samples we obtained from the NICU. Sets of 8 samples (total of 3 sets) were pooled. For both experiments each pool of milk was divided into two fractions, one fraction was immediately processed (MOM) for EV enrichment (MOM EV ), while the other half was pasteurized to mimic DHM (P-MOM) to obtain PMOM EV , following the HMBANA protocol (HMBANA 2020). Briefly, milk was heated at constant temperature 65 °C for 30 min and cooled down immediately after in an ice bath. After pasteurization milk was further processed for EV enrichment as follows. MOM and PMOM fractions were centrifuged at 2246 g for 15 min at 4 °C to remove the fat, following a 45 min centrifugation at 15900 g to eliminate cell debris. Milk supernatants were further filtered sequentially using nitrocellulose filters of 11 um, 6 um, 2.5 um, 0.45 um and 0.2 um. After filtration, EVs were enriched by ultracentrifugation at 207,888 g for 2 h. The EVs were washed twice with PBS, quantified, and stored in aliquots at − 80 °C.
EV physical characterization and quantification. Nanoparticle tracking analysis (NTA) using a NanoSight NS300 (Malvern Instruments Ltd, Malvern, UK) was utilized to quantify the EVs as well as to determine its size distribution. Videos were recorded for 60 s (five times), with the camera level at 15, and analyzed with NTA software 4.3 (Malvern instruments Ltd, Malvern, UK). An average yield of 10 12 exosomes per uL was obtained. Dynamic light scattering was performed to measure the zeta potential of the EV suspensions using a Zetasizer ultra particle analyzer (Malvern Instruments Ltd, Malvern, UK). The samples were diluted 1:1000 with distilled water. The measurements were conducted in biological and technical triplicates at 25 °C.
RNA extraction, sequencing, and data analysis. RNA extractions were performed from three pools of MOM EV and PMOM EV samples using mirVana™ miRNA Isolation Kit (Invitrogen), with small RNA enrichment from total RNA according to manufacturer's instructions. Library construction with fragments around 20-75 bp in length and sequencing using Ion Torrent sequencing platform was performed by PrimBio (PrimBio Research Institute LLC, Exton, PA). FastQC was used to filter high quality reads 58 . For the data analysis Kallisto v0.46.1 59 was used to create an index and map the high-quality raw reads using a reference transcriptome for no coding RNA through EnsDb.Hsapiens.v86 2.99.0 package in RStudio 60,61 . Differential abundance analysis of miRNA was performed using limma 3.52.4, edgeR 3.38.4 and SVA 3.44.0 packages in RStudio [62][63][64] . All the graphics were generated using ggplot2 3.3.6 R package 65 . Protein extractions and Western Blot. Aliquots normalized to the same particle concentration of MOM EV and PMOM EV were used for protein extraction and quantification using the Pierce™ BCA Protein Assay Kit (Thermo Scientific, Rockford, IL). Briefly, total proteins were extracted from EVs using Radio Immunoprecipitation Assay Buffer (RIPA) containing 150 mM NaCl, 50 mM Tris (pH 8), 1% Triton X-100, and 0.1% sodium dodecyl sulfate (SDS), with Halt™ protease inhibitor cocktail (Thermo Fisher, Waltham, MA, USA). The EVs homogenates were centrifuged at 12,000 g for 10 min, at 4 °C and the protein concentration was measured following the manufacturer instruction.
Lipoproteins HDL (High-Density lipoprotein) and LDL (Low-Density lipoprotein) were purified from human milk EVs suspensions using LDL/VLDL and HDL purification kit (STA-608 ultracentrifugation free) following manufacturer protocol (Cell Biolabs, INC). Lipoproteins were solubilized in equal volume of PBS as EVs suspensions.
qRT-PCR and mRNA expression. RNA was isolated from cell lines using RNeasy® Mini Kit following the manufacturer's protocol (QIAGEN, Germantown, MD). DNA was removed by treatment with DNase (QIAGEN, Germantown, MD) according to the manufacturer's protocol. RNA quality was monitored on 1% agarose gels, and RNA quantification was performed using Thermo Scientific Nanodrop One Microvolume UV-vis spectrophotometer (Thermo Fisher Scientific, Grand Island, NY). qRT-PCR was performed as described 66 . Primer sequences used to determine relative transcript abundance are listed in Suppl. Table 3.