Formalin was purchased from VWR international (West Chester, PA). Recombinant human erythropoietin (rhEPO) was purchased from Amgen Biologicals (Thousand Oaks, CA).

The animals were kept under standardized temperature, humidity, and lighting conditions, with free access to water and food. Postnatal day 3 (P3) rat pups were subjected to 5% formalin (10 μl) subcutaneous (s.c.) injections into the four paws. rhEPO was administered 30 min before formalin injection in the EPO treated group. Some animals were injected once and sacrificed 4 or 12 hours later; while others were injected for three consecutive days and sacrificed on postnatal day 6 or 21 (P6 or P21). Pups were returned to their mothers immediately after each injection. The health condition of each pup was monitored each day for feeding behaviors and signs of feet swelling, skin infection, and pain development. Animals with skin infection, excessive swelling and pain were sacrificed without further testing. Mild paw swelling and skin redness recovered fully within 3 days after injection. After sacrifice, brain and dorsal root ganglia (DRG) tissues were collected for immunostaining and/or Western blot analysis. All studies were approved by the Institutional Animal Care and Use Committee (IACUC).

Under isofluorane anesthesia, animals were transcardially perfused with 10% formalin. Brain coronal sections were sliced at 20 μm thickness using a cryostat vibratome (Ultapro 5000, St louis, MO, USA) and were mounted on gelatin coated slides. Fluoro-Jade C (FJC) staining 29 , a valid and reliable stain for degenerating neurons, was conducted as reported earlier with some modifications. In brief, tissue sections were dried for 15 minutes. After three PBS washes, the sections were immersed in 0.8% NaOH in 80% ethanol followed by 70% ethanol and water washes, respectively. The sections were then immersed in 0.06% KMnO₄, followed by 0.0002% FJC. The section were thereafter washed, dried and mounted for microscopy using the FITC channel on an upright Olympus fluorescence microscope.

TUNEL staining was performed using a commercial kit (DeadEnd™ Fluorometric TUNEL system, Promega, Madison, WI, USA) to label dying and dead cells in the brain. The instructions were followed as given. In brief, brain sections were placed in equilibration buffer and incubated with nucleotide mix and rTdT enzyme at 37°C for 1 hr. The reaction was stopped with 2 × SSC. Results were visualized using the FITC channel on an upright Olympus fluorescence microscope.

Cell count was performed following the principles of design based stereology. Systematic random sampling was employed to ensure accurate and non-redundant cell counting 30 . Every section under analysis was at least a 100 μm away from the next. A total of 6 20-μm thick sections spanning the entire regions of interest were randomly selected for cell counting from each animal. Counting was performed on 6 non-overlapping randomly selected 20 × fields per section. Sections from different animals represent the same area in the anterior posterior direction.

Brain tissue samples were collected, proteins were extracted with lysis buffer, and protein concentration was measured with a BCA assay (Pierce, Rock Ford, IL, USA). 30 μg protein samples were electrophoresed on 6 to 15% gradient gels in the presence of sodium dodecyl sulfate-polyacrylamide in a Hoefer Mini-Gel system (Amersham Biosciences, Piscataway, NJ, USA) and transferred in the Hoefer Transfer Tank (Amersham Biosciences, Piscataway, NJ, USA) to a polyvinylidene difluoride membrane (BioRad, Hercules, CA, USA). Membranes were blocked in 7% milk (Tris-buffered saline containing 0.1% Tween-20, pH 7.6%, 7% milk) at room temperature for 1 h and incubated overnight at 4°C with rabbit polyclonal anti-caspase-3 (1:500, Cell Signaling, Boston, MA) and rabbit anti-apoptosis inducing factor AIF (1:1000, Cell Signaling, Boston, MA) antibodies. Mouse anti-actin antibody (1:5000, Sigma, St Louis, MO, USA) was used to detect β-actin, the protein loading control. The blots were washed in 0.5% Tris-buffered saline containing 0.1% Tween-20 and incubated with alkaline phosphatase conjugated anti-rabbit or anti-mouse (1:2000, Promega, Madison, WI, USA) antibodies for 2 hrs at room temperature. Finally, membranes were washed with Tris-buffered saline containing 0.1% Tween-20 followed by three washes with Tris-buffered saline. The signal was detected by the addition of bromo-chloro-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT) solution (Sigma, St Louis, MO, USA), and quantified and analyzed by the imaging software Photoshop Professional (Adobes Photoshop CS 8.0, San Jose, CA, USA). This whole procedure was repeated three times with new samples collected from different animals.

Total RNA was extracted from the somatosensory cortex (4 and 12 hours after single injection and after 3 days' consecutive injections at P6 and P21) and dorsal root ganglia (12 hours after single injection) using the TRIzol reagent (Invitrogen Inc, Carlsbad, CA). Reverse transcription was performed with 2 μg total RNA using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems, CA, USA). SYBR green qRT-PCR was used to assess the relative levels of our target genes using the Applied Biosystems StepOnePlus machine. The primers used were: TNF-α: F-primer: CCCCATTACTCTGACCCCTT; R-primer: TTGTTGGAAATTCTGAGCCC; Il-6: F-primer: GCCCTTCAGGAACAGCTATG; R-primer: CCGGACTTGTGAAGTAGGGA; Il-1β: F-primer: CATCTTTGAAGAAGAGCCCG; R-primer: AGCTTTCAGCTCACATGGGT; substance P (SP): F-primer: TGACCAAATCAAGGAGGCAAT; R-primer: GGGTCTTCGGGCGATTCT; Calcitonin gene related peptide (CGRP): F-primer: AAGTTCTCCCCTTTCCTGGTTGTCA; R-primer: TGGTGGGCACAAAGTTGTCCTTCAC;

Fold change was calculated by the delta (delta Ct) method using the 18S ribosomal RNA subunit amplification as the internal control. All experiments were repeated three times using distinct animals by two different investigators.

To evaluate the behavioral changes in the experimental and control groups, we utilized the following behavioral tests.

Cerebral blood flow was measured using laser-Doppler flowmetry (PeriFlux System 5000 - PF5010 LDPM unit, Perimed, Stockholm, Sweden). The tip of a flexible 0.5 fiber optic probe was pointed over the somatosensory cortex. Measurements were taken before and after the last injection of formalin on postnatal day 5 during the third injection session. The final value is the ratio of blood flow after to before injections and it is an average of five consecutive readings.

All analyses were performed using GraphPad Prism 4.0 statistical software (GraphPad Software, Inc., La Jolla, CA). Multiple comparisons were performed by one- or two-way analysis of variance (ANOVA) followed by Bonferroni's post hoc analysis. Single comparisons were performed using Student's t-test. Changes were considered significant if the p-value was less than 0.05. Mean values were reported with the standard error of the mean (SEM).

The subcutaneous formalin administration into the paws of rodents is a well established nociceptive insult used in peripheral inflammatory pain models 31 32 . P3 neonatal rats were randomized to four groups: saline injection controls, rhEPO injection controls, formalin injection group and formalin coupled to intraperitoneal (i.p.) injection of rhEPO (5,000 U/kg). It is estimated that the rat brain at postnatal days 1-10 is equivalent to a third trimester human brain and that rat neurodevelopment at postnatal day 7 is equivalent to that of the human brain at birth 33 34 . The postnatal day 3 was selected to mimic premature birth in humans. rhEPO was administered 30 min before each formalin injection in the formalin plus EPO treatment group. Formalin (5%, 10 μl) was subcutaneously injected into the four paws, starting on P3. Animals sacrificed at 4 and 12 hours received only one treatment with saline control, rhEPO alone, formalin alone or formalin plus rhEPO, respectively. Animals sacrificed at P6 (72 hrs) and P21 received injections from P3 to P5 once a day in each paw. Rat pups receiving formalin injections showed skin redness and moderate swelling at injection sites. These changes fully reversed in 3 days. We regard the employed formalin model as a severe and subacute inflammatory pain model, suitable for generating noticeable short- and long-term alterations.

As a control study for identifying the effect of rhEPO on cytokine expression and animal behavior, we injected rhEPO alone in P3 neonatal rats and inspected cytokine expression 12 hrs later, and cytokine expression and behavior at P21. Injection of rhEPO into P3 neonatal rats did not affect IL-6, TNF-α and IL-1β transcription levels 12 hrs later (Figure 1A to 1C). Moreover, rhEPO, injected three times in neonatal rats at P3, P4 and P5, did not affect transcription of inflammatory cytokines (IL-6, TNF-α and IL-1β) examined at P21 (Figure 1D to 1F). In addition, the behavior of P21 juvenile rats (injected with rhEPO at P3, P4 and P5) was not affected in the defensive-withdrawal, hot-plate or Rota-rod tests (Figure 1G to 1I).

TUNEL staining was used to investigate the extent of cell death in the brain. TUNEL identifies fragmented DNA in all types of dying cells 35 36 . In addition, we used Fluoro-Jade C (FJC) to specifically identify dying neurons 29 . The analysis of data from TUNEL staining after 3 sessions of formalin injection (72 hrs) revealed that the peripheral inflammatory nociceptive insult caused widespread cell death in the cortex and hypothalamus (Figure 2A, C and 2E), whereas rhEPO co-treatment effectively reduced cell death in these areas (Figure 2B, D, and 2F). Counting of TUNEL-positive cells in the cortex, hippocampus and hypothalamus showed a significant effect of rhEPO in reducing cell death after repeated formalin injections (Figure 3E). In the hippocampus and hypothalamus, rhEPO treatment showed significant protection. TUNEL positive cells were reduced from 303 ± 44/field in the formalin alone rats to 227 ± 59/field, and from 278 ± 50/field to 224 ± 32/field in these two regions, respectively (n = 6 and p < 0.05 in both analyses). The number of TUNEL-positive cells in the cortex of rhEPO-treated rats was 145 ± 24 as compared to 170 ± 37/field in the formalin alone group (however, p > 0.05, n = 6 rats in each group). The more specific neuronal stain FJC further verified that treatment with rhEPO dramatically reduced neuronal cell death in the cortex, hippocampus (Figure 2G to 2I) and hypothalamus after a single (4 and 12 hrs) or 3 formalin injections (72 hrs) (Figure 3A to 3C). For example, 12 hrs after a single injection, rhEPO reduced FJC-positive cells from 408 ± 41/field to 148 ± 12/field in the cortex, from 1136 ± 71/field to 368 ± 23/field in the hippocampus, and from 548 ± 50/field to 182 ± 21/field in the hypothalamus (p < 0.05 vs. saline control in all three comparisons; n = 6 in each group) (Figure 3B). At age P21, the number of dying neurons was similar in the control, formalin injected and EPO treated groups and they were significantly lower than the earlier time points (Figure 3D).

Immature neurons are more vulnerable to apoptotic insults 37 . We tested the hypothesis that formalin injection triggers the activation of apoptotic mechanisms in the neonatal brain. To check for this possibility, we performed Western blot analysis on somatosensory cortical tissues collected after 3 formalin injections in 3 days and probed for the levels of caspase-3 and apoptosis-inducing factor (AIF). Formalin injections led to a significant rise in the levels of cleaved caspase-3 compared to saline group (1.48 ± 0.13 vs. 0.97 ± 0.17 folds; p < 0.05), whereas rhEPO co-treatment strongly prevented caspase-3 activation (0.84 ± 0.16 vs. 1.48 ± 0.13 folds; p < 0.05) (Figure 3F). The formalin insult did not change the cytosolic levels of AIF, which represents the caspase-independent apoptosis 38 .

To further characterize the mechanisms by which formalin caused widespread cell death in the brain, we aimed to identify the transcriptional levels of some major inflammatory cytokines involved in the pathogenesis of inflammatory pain. To measure transcription, we used real-time PCR to follow mRNA production of IL-6, IL-1β, and TNF-α in the somatosensory cortex and substance P (SP) and calcitonin-gene-related peptide (CGRP) in the dorsal root ganglia (DRG) corresponding to the upper and lower limbs. In the somatosensory cortex, formalin injection significantly increased Il-6 at the 4 hrs time point (19.0 ± 3.0 folds increase; p < 0.05) (Figure 4E). TNF-α level was significantly enhanced at the 4 hr (11.1 ± 2.7 folds) and 72 hr (13.0 ± 3.5 folds) time points (p < 0.05 in both comparisons) (Figure 4I and 4K). IL-1β level increased 2.3 ± 0.5 folds (p < 0.05 compared to controls) (Figure 4B) at the 12 hrs time point. On the other hand, formalin injection reduced IL-1β levels at 4 and 72 hrs after injection (Figure 4A and 4C). rhEPO treatment effectively prevented the increases in IL-6 and TNF-α levels at 4 and 72 hrs, respectively (Figure 4E and 4K). However, rhEPO did not reduce the levels of TNF-α 4 hrs after formalin injection. TNF-α transcription level was still high at this early time point (Figure 4I). In addition, there was no significant difference in the transcriptional levels of IL-1β, IL-6, or TNF-α when measured at age P21 after 3 injection sessions (Figure 4D, H and 4L). Moreover, we detected a strong trend of decreased SP level (0.31 ± 0.20 folds vs. the control; but p > 0.05) and a decreased CGRP level (0.09 ± 0.08 folds vs. the control; p < 0.05) 12 hrs after a single formalin injection. Treatment with rhEPO significantly increased the mRNA expression of SP (2.2 ± 0.2 folds vs. formalin alone; p < 0.05) and CGRP (1.9 ± 0.4 folds vs. formalin alone; p < 0.05) (Figure 4M and 4N).

A major complication of early exposure to painful experiences is altered pain sensation in adult life 39 3 . This fact prompted us to test pain sensitivities in 21-day old juvenile rats after they have received three injection sessions of saline, formalin alone or formalin plus rhEPO in infancy (P3 to P5). We noticed that rats which received repetitive formalin injections had reduced pain thresholds in the hot-plate test (9.9 ± 0.9 sec vs. 17.1 ± 2.8 sec; p < 0.05). Interestingly, there was no difference in their pain thresholds in the tail-flick test (1.5 ± 0.2 sec vs. 1.5 ± 0.3 sec; p > 0.05) (Figure 5A and 5B), suggesting that the peripheral inflammatory stimuli selectively affect supra-spinal but not spinal pain pathways. Treatment with rhEPO prevented the hyperalgesia in the hot-plate test (Figure 5A), which is consistent with the rhEPO effect in L5 spinal nerve transection neuropathic pain model 40 .

To check if the altered pain sensitivity is associated with increased brain activity during the painful stimuli, we measured local cerebral blood flow (LCBF) in the brain 5 min before and immediately after formalin injection on the third and last day of injections in animals that have received three injection sessions (postnatal day 5). Formalin injection significantly increased blood flow in the right somatosensory cortex; whereas treatment with rhEPO restored normal flow (Figure 5C and 5D). Same results were obtained from the left hemispheres (Figure 5E). These data suggested that peripheral inflammatory pain causes a state of hypersensitivity/hyperactivity in the brain and rhEPO was able to restore normal brain activity levels upon the painful stimuli.

Behavioral disturbances and psychiatric disorders are major complications that stem from early painful experiences. Defensive-withdrawal test evaluates exploratory behavior in rats. After 3 sessions of formalin injections in the neonatal period, we tested exploratory behavior in the juvenile 21-day old rats. Formalin injections increased the time spent in the dark chamber (456 ± 95 sec vs. 10 ± 3 sec in control rats; p < 0.05). rhEPO treatment corrected this abnormal behavior back to the normal level (17 ± 8 sec; p > 0.05 compared with controls) (Figure 6A). To rule out the possibility that the formalin insult impaired locomotor functions, we tested locomotion using the Rota-rod and open field tests in the P21 rats. There were no statistically significant differences on either test among the three groups (Figure 6B and 6C).

In the long-term experiments, we tested the effect of formalin injections on brain and body weights in the P21 rats of the three experimental groups. Rats that received formalin injections during their neonatal period showed significantly less body (25.8 ± 1.0 g vs. 39.7 ± 0.6 g; p < 0.05) and brain (1.1 ± 0.02 g vs. 1.3 ± 0.03 g, p < 0.05) weights compared to the control group. These decreases in body and brain weights were not seen in rats that also received rhEPO. Their body and brain weights remained within the normal ranges (Figure 7A and 7B).