FAIM-L regulation of XIAP degradation modulates Synaptic Long-Term Depression and Axon Degeneration

Caspases have recently emerged as key regulators of axonal pruning and degeneration and of long-term depression (LTD), a long-lasting form of synaptic plasticity. However, the mechanism underlying these functions remains unclear. In this context, XIAP has been shown to modulate these processes. The neuron-specific form of FAIM protein (FAIM-L) is a death receptor antagonist that stabilizes XIAP protein levels, thus preventing death receptor-induced neuronal apoptosis. Here we show that FAIM-L modulates synaptic transmission, prevents chemical-LTD induction in hippocampal neurons, and thwarts axon degeneration after nerve growth factor (NGF) withdrawal. Additionally, we demonstrate that the participation of FAIM-L in these two processes is dependent on its capacity to stabilize XIAP protein levels. Our data reveal FAIM-L as a regulator of axonal degeneration and synaptic plasticity.

Data represent mean ± SEM and One-way ANOVA followed by Tukey's multiple comparison post-hoc test was used to calculate significant levels between the indicated groups. *p < 0.05; **p < 0.01; ***P < 0.001; ****p < 0.0001. (C) Hippocampal neurons were infected with lentiviral vectors as indicated. Protein levels of XIAP and FAIM-L were determined by Western blot 48 h after infection with overexpressing and silencing vectors. Data are represented as the mean ± standard error of the mean (SEM) of three independent experiments. One-way ANOVA followed by Tukey's multiple comparison post-hoc test was used to calculate significant levels between the indicated groups. *p < 0.05; **p < 0.01.  13 and once the axon is degenerating, XIAP is also responsible for restricting caspase activation to this specific subcellular compartment 4 . XIAP has also been linked to the control of synaptic plasticity, thus its regulation of caspase-3, -7 and -9 activation modulates long-term depression (LTD)-induced AMPA receptor (AMPAR) internalization 6 . However, the contribution of FAIM-L-mediated XIAP stabilization to these non-apoptotic functions of XIAP has not been explored.
Here we examined the relevance of FAIM-L-mediated XIAP stabilization in two critical non-apoptotic processes, namely axon degeneration and synaptic plasticity. We report that the precise regulation of XIAP protein levels by FAIM-L affects the synaptic plasticity of hippocampal neurons exposed to NMDA-induced LTD, as well as the axon degeneration of dorsal root ganglion neurons (DRGs) induced by NGF withdrawal. Our data suggest that FAIM-L plays a key role in the maintenance of life-long neuronal plasticity by regulating axon degeneration and synaptic plasticity. Mitochondria engagement, and caspase-3 activation and its modulation by XIAP participate in LTD regulation, a crucial physiological process for CNS development and homeostasis [6][7][8] . Our previous data revealed the capacity of the DR antagonist, FAIM-L, to stabilize XIAP levels, and thus, to protect cortical neurons from FasL-induced cell death 2 . Therefore, we first assessed the capacity of FAIM-L to regulate XIAP levels in hippocampal neurons. Cultured neurons were transduced with FAIM-L or empty EGFP-tagged lentiviral vectors for overexpression and shRNA-FAIM-L or shRNA-scrambled for gene silencing. XIAP expression was assessed by immunocytochemistry and Western blot. The overexpression or silencing of FAIM-L in hippocampal neurons resulted in the up-regulation or the abolishment of XIAP expression, respectively ( Fig. 1A-C).

FAIM-L regulates AMPAR internalization after NMDA-induced LTD in hippocampal neurons.
Li and collaborators reported that XIAP and other members of the intrinsic apoptotic pathway regulate the internalization of the AMPAR subunit GluA2 after chemical-LTD induction. They proved that XIAP overexpression blocks NMDA-induced GluA2 internalization not only in hippocampal cultures, but also in CA1 organotypic slice cultures, where XIAP overexpression abrogated the LTD induced by low frequency stimulation 6 . Since we observed that FAIM-L regulates XIAP protein stability in hippocampal neurons, we next assessed its relevance in LTD. To address this issue, we used an in vitro model of LTD (chemical-LTD) that consists of NMDA treatment of hippocampal neurons for 15 min. The induction of LTD is monitored by the internalization of AMPAR subunits, such as GluA1 and GluA2. AMPAR endocytosis was measured in dissociated hippocampal neurons by means of an "antibody feeding" internalization assay 14 . Hippocampal neurons were infected with FAIM-L or empty EGFP-Tagged lentiviruses for overexpression, together with shRNA-XIAP, shRNA-FAIM-L or shRNA-scrambled for gene silencing (Figs 2 and 3). At 15-18 DIV, cultures were treated with 50 μ M NMDA for 15 min at 37 °C to stimulate chemical-LTD-induced GluA2 internalization. In control conditions (see + shRNA-scrambled, + EMPTY-EGFP; Fig. 2A,B,I), the internalization was reduced when FAIM-L was overexpressed, as no significant differences were observed when comparing GluA2 internalized levels after treatment with control medium or NMDA in this condition (see + shRNA-scrambled, + FAIML-EGFP; Fig. 2C,D,I). In agreement with the observations of Li et al. 6 , we found that NMDA-induced GluA2 internalization was greatly increased when XIAP was knocked down using specific shRNA (see + shRNA-XIAP, + EMPTY-EGFP; Fig. 2E,F,I). Moreover, the reduction of GluA2 internalization caused by FAIM-L overexpression was not detected after XIAP silencing, thus further strengthening the notion that FAIM-L exerts its effect through the stabilization of XIAP protein (see + shRNA-XIAP, + FAIM-L-EGFP; Fig. 2G,H,I). The efficacy of shRNA-XIAP on XIAP expression was tested in Fig. 2J.
Finally, we tested the participation of endogenous FAIM-L in regulating the internalization of AMPAR subunit GluA2. With this aim, we examined the levels of GluA2 internalization after NMDA treatment in neurons in which FAIM-L was silenced. Strikingly, GluA2 significantly enhanced internalization even in the absence of NMDA treatment, thereby pointing to the participation of FAIM-L in the regulation of AMPAR trafficking in basal conditions (see + shRNA-FAIM-L; Fig. 3C,E). Additionally, as expected, the induction of GluA2 internalization by NMDA was higher than in the control condition ( Fig. 3A-E), resembling a shRNA-XIAP-like phenotype (Fig. 2E,F,I). These observations strongly support the notion that FAIM-L contributes to the regulation of synaptic plasticity.
remaining GluA2 (third column, green in merge), internalized GluA2 (fourth column, red in merge), and merge (fifth column). Individual channels are shown in gray scale. Images in columns 2, 3, 4 and 5 represent magnifications from selected areas of the first columns. The scale bar represents 20 μ m. (I) Shows quantitation of internalization index for the experiment represented in (A-H) integrated fluorescence intensity of internalized GluA2/integrated fluorescence intensity of surface-remaining GluA2. Results were not normalized to untreated cells (− NMDA) 6 . As a consequence, A.U have no dimensions. N = 55 to 67 neurons from 3 independent experiments, for each group. Data represent mean ± SEM and and One-way ANOVA test followed by Tukey's multiple comparison post-hoc test was used to calculate significant levels between the indicated groups. *p < 0.05; ***P < 0.001; ****p < 0.0001. (J) control experiment to demonstrate that the used XIAP-shRNA efficiently down-regulates the expression of its target protein. Hippocampal neurons were infected with lentivirus containing the scramble-shRNA vectors (upper panels) or XIAP-shRNA vectors (lower panels) and 72 h later they were fixed and immune-stained against XIAP. In lower panels, it can be clearly appreciated that only neurons infected with XIAP-shRNA showed a clear reduction of XIAP protein levels. However, the XIAP protein levels are not affected in non-infected neurons or neurons infected with scramble-shRNA.
Our data suggest this DR antagonist modulates synaptic transmission. Li and co-workers 6 discarded the participation of caspase activation in synaptic transmission, and therefore the mechanism through which FAIM-L exerts its function in this context remains elusive. Nevertheless, these observations coincide with the effect on AMPAR internalization observed in the FAIM-L-transduced hippocampal neurons, where the over-expression of this protein resulted in greater internalized AMPAR in non-stimulated neurons (Fig. 2C-I).
Overexpression of FAIM-L blocks LTD. To study the role of FAIM-L in long-term depression LTD in pyramidal neurons (Fig. 4D,E) we measured mEPSCs amplitude before and after a treatment with NMDA (50 μ M during 5 min), a well-established protocol to induce chemical LTD (chem-LTD) either in the CA1 region of hippocampus and in hippocampal neurons in culture 15,16 . Miniature events were recorded in 5 min periods for a total of 20 min after NMDA stimulation. NMDA application produced a significant LTD as seen as a decrease in mEPSC amplitude as early as 10  FAIM-L regulates axonal degeneration. During development, neurotrophins are required for cell survival, and they also contribute to neurite growth and maintenance. The action of these molecules results in an overproduction of neural connections that are later removed to form the correct patterns of connectivity. Both neuronal loss and selective developmental axonal degeneration contribute to the adjustment of neuronal connections. Although caspases mediate neuronal apoptosis, they also participate in axonal degeneration 3 . XIAP regulates caspase activation and, as expected, also participates in this process 13 . Since we observed that FAIM-L-promoted XIAP protein stabilization participated in the regulation of a non-apoptotic function of XIAP and caspases such as the LTD, we sought to analyze whether FAIM-L-mediated XIAP stabilization is also involved in axonal degeneration.
First, to address this question we took advantage of a model of axonal degeneration in DRG explant cultures. We dissected out the explants and infected them with the lentiviral vectors for EMPTY-EGFP or FAIM-L-EGFP. Axons were left to grow for 2 DIV and then subjected to NGF deprivation (Fig. 5A). Strikingly, FAIM-L levels in the explants were reduced under these conditions (Fig. 5B). Moreover, FAIM-L overexpression partially protected the explant axons from degeneration after 8-24 h of NGF withdrawal (Fig. 5A).
Many lines of evidence have been published in the field reporting that caspase-3 is involved in axonal degeneration related to axonal pruning during development [3][4][5] . Indeed, XIAP regulates this process through caspase-3 inhibition. Thus, we addressed whether FAIM-L overexpression interferes with caspase-3 activation during axonal degeneration. Infected explants were deprived of NGF for 8 and 24 h and caspase-3 activation was assessed by Westrn blot. FAIM-L overexpression reduced the generation of the caspase-3 active form p17 at 8 h post-deprivation (Fig. 5B). It has been reported that developmental prompted axonal degeneration involves the activation of calpain by caspase-3 degradation of its inhibitor calpastatin. Therefore, one of the characteristic features of pruning is the appearance of degraded substrates of calpains such as neurofilament 66 (NF-66) 5 . Consistent with the activation of caspase-3 observed in the explants exposed to NGF withdrawal, we detected fragmented forms of NF-66. In this context, the overexpression of FAIM-L induced a reduction in the caspase-3 active fragment at 8 h and a decrease in NF-66 processing at 24 h compared to EMPTY-EGFP-infected explants (Fig. 5B).

Endogenous XIAP is required for FAIM-L regulation of axonal degeneration.
To further characterize FAIM-L modulation of axonal degeneration, we used Campenot chamber cultures. In these cultures, neurons are seeded in compartmented (Campenot) chambers that allow the establishment of two separate fluid environments. Briefly, DRGs are placed in a central chamber containing NGF, and axons grow inside the lateral chambers (Fig. 6A). Due to limited fluid exchange between the chambers, it is possible to achieve local neurotrophin deprivation that affects only axons, while cell bodies continue to be sustained by NGF (Fig. 6A-D).
To determine the role of the FAIM-L/XIAP axis in axonal pruning, DRG neuron bodies were transfected with XIAP-shRNA or scrambled-shRNA and FAIM-L-EGFP or EMPTY-EGFP overexpression constructs (Figs 6E and 7A-E). In agreement with what we previously described in cortical neurons 2 and the results reported above for hippocampal neurons (Fig. 1B), FAIM-L overexpression promoted the stabilization of XIAP protein levels in DRG neurons (Fig. 6E). When DRG underwent local NGF deprivation, FAIM-L overexpression prevented the and NF-66 cleaved fragment relative levels quantification to Histone H3, total caspase-3 and total NF-66, respectively, from three independent experiments in the indicated conditions. Data are represented as the mean ± standard error of the mean (SEM). Two-way ANOVA test followed by Bonferroni post-hoc test was used to calculate significant levels between the indicated groups. *p < 0.05.   Histogram representing active caspase-3 levels relative to total caspase-3 and NF-66 cleaved fragment relative to total NF-66 in each condition 24 h after NGF withdrawal. Data are represented as the mean ± SEM of three independent experiments. One-way ANOVA with Tukey's multiple comparison post-hoc test was used to calculate significant levels between the indicated groups. *p < 0.05. **p < 0.01.
Scientific RepoRts | 6:35775 | DOI: 10.1038/srep35775 axonal degeneration observed after 24 h of NGF withdrawal in control DRG neurons (Fig. 7A,B,E). To test the relevance of FAIM-L-induced XIAP protein stabilization for the capacity of FAIM-L to prevent axon degeneration, we performed the same experiment in the presence of XIAP shRNA (Fig. 7C-E). XIAP down-regulation abolished FAIM-L blockade of axon degeneration, thereby demonstrating that XIAP participates in the effect of FAIM-L on the regulation of axonal degeneration (Fig. 7E). Accordingly, the removal of endogenous FAIM-L levels resulted in faster degeneration when axons were exposed to NGF withdrawal (Fig. 8A). Therefore, our observations in the Campenot chambers support the idea that FAIM-L modulates axonal degeneration in physiological conditions by stabilizing XIAP.
Finally, we determined the status of caspase-3 activation and NF-66 degradation by Western blot. The presence of FAIM-L detected after 24 h of NGF depletion resulted in a reduction of the p17 fragment of caspase-3 and a blockade of NF-66 degradation (Fig. 8B,C). Consistent with our observations regarding axon degeneration, XIAP silencing was sufficient to block FAIM-L inhibition of caspase-3 activation and cleavage of NF-66 (Fig. 8B). These findings further support the notion that FAIM-L exerts its effect by stabilizing endogenous XIAP levels.
All together our data corroborate that the FAIM-L/XIAP/caspase axis is involved in non-apoptotic, neuronal physiological processes, such as axonal pruning and synaptic plasticity.

Discussion
Here we provide evidence that the DR antagonist FAIM-L regulates axon degeneration and synaptic plasticity by controlling XIAP protein levels. We recently characterized the mechanism underlying the role of the FAIM-L/ XIAP axis in neuronal protection against apoptosis. We described that FAIM-L interacts with XIAP through its IAP-binding motif and impairs its auto-ubiquitinylation and proteasomal degradation, subsequently inhibiting caspase activation. The inhibition of XIAP degradation protects neurons against Fas-induced apoptosis 2 . In the present work, we report other functions of the FAIM-L/XIAP axis in a neuronal physiological context that can be added to the previously described role of apoptosis control.
Caspases have been widely related to the resolution of apoptotic cell death 17 . However, these molecules are also activated in neuronal processes that are no longer functional and are marked for elimination 18,19 . In this context, it is reasonable to hypothesize that caspases exert functions that do not depend on cell death induction. It is particularly relevant that FAIM-L, an antagonist of apoptosis, is selectively expressed in neurons. As discussed above, caspases are involved in axonal pruning [3][4][5] and in the regulation of synaptic plasticity [6][7][8] .
Mitochondria are well-suited to play a role in synaptic plasticity as this phenomenon modulates their distribution, morphology and motility 15 . Mitochondria in dendrites take up calcium after synaptic stimulation 16 , and this uptake in turn promotes the release of pro-apoptotic factors from these organelles 20,21 . Thus, mitochondria are crucial for LTD induction. Li and colleagues reported that the induction of LTD caused by the reduction of AMPA receptor exposure in the membrane is regulated by caspase activity 6 . From the first identification of caspase involvement in this process, most apoptotic members of the intrinsic pathway have been found to correlate with synapse weakening [6][7][8] . In contrast, proteins that negatively regulate caspase activation, such as XIAP, impede LTD 6 . Our results support all these findings since the silencing of FAIM-L expression in hippocampal neurons dramatically increased the induction of GluA2 internalization in response to NMDA treatment, as observed in XIAP-silenced cells. In agreement with these data, when XIAP protein levels were stabilized by overexpressing FAIM-L, hippocampal neurons did not exhibit significant GluA2 internalization levels after NMDA treatment compared to non-treated neurons and, accordingly, they did not show LTD induction after NMDA treatment.
Nevertheless, our data not only describes the involvement of FAIM-L in the regulation of synaptic plasticity process LTD, but the participation of this DR antagonist in synaptic transmission. We only could detect that FAIM-L overexpression made the neurons prone to a more excited phenotype, as there was an increase in the amplitude of mEPCS, thereby indicating a perturbation of the correct function of the pre-and the post-synaptic synapse in this condition (Fig. 4A,B,C). This is the first time that a DR antagonist has been linked to the modulation of synaptic transmission. The involvement of FAIM-L modulation of XIAP in the mechanism underlying this process is still unknown. There is no clear evidence pointing to XIAP as a modulator of synaptic transmission, and the mediation of caspases in this process has been discarded 6 . Thus, more experiments are required to elucidate whether FAIM-L performs this function through the stabilization of XIAP.
XIAP interacts and inhibits the effector caspases-3 and -7 and the initiator caspase-9 9-12 . The physiological relevance of XIAP regulation of apoptosis has been related to the modulation of innate immunity, inflammation, and oncogenesis 22,23 ; however, there is little evidence of its relevance in the CNS. During the development of the nervous system in Drosophila, the XIAP homolog DIAP-1 plays a key role in the regulation of caspase activity. DRONC caspase is constitutively active in fly neurons, and therefore DIAP1 is required to restrain its activity during neuronal growth. The inactivation of DIAP-1-mediated DRONC inhibition via DIAP-1 proteasomal degradation is crucial for the reshaping of dendritic arbors on ddAc sensory neurons during morphogenesis 24 . Unsain and collaborators demonstrated that the XIAP-caspase loop that regulates morphogenesis is phylogenetically conserved from flies to mammals 13 . Given our results, we propose that FAIM-L modulates this loop through the stabilization of XIAP. We observed that the overexpression of this DR antagonist not only reduces the axonal degeneration induced by NGF withdrawal but also impairs caspase-3 activation, which is consistent with the inhibition of XIAP degradation. Accordingly, cleavage of NF-66, a target of caspase-3 activation, is also subjected to interference (Figs 5 and 8).
As mentioned above, XIAP regulation of caspase activity in degenerating axons has been elegantly reported by Dr. Barker's and Dr. Deshmukh's labs 4,13 . Although both studies report higher caspase-3 activation and axon degeneration in XIAP KO mice, only one of them shows a decrease of XIAP levels after NGF withdrawal 13 . Moreover, in the study where the participation of caspase-3 in axon degeneration was stated, they did not observe any changes in XIAP axonal levels after NGF deprivation 3 . In this regard, we have observed low or not evident XIAP down regulation after NGF withdrawal in DRG explant cultures; therefore, we have been unable to study if FAIM-L overexpression modulates XIAP levels in this context. Nevertheless, we observed a reduction in caspase-3 activation and a reduction in NF-66 degradation when we overexpressed FAIM-L. Taking into account these results, the observation that FAIM-L levels are decreased after NGF depletion, and the data shown in Fig. 8A, where we demonstrate that FAIM-L down-regulation by shRNA induces a faster axon degeneration, we believe that FAIM-L participates actively in this process.
The observation that NGF deprivation of DRG explants is accompanied by a downregulation of FAIM-L suggests that the modulation of the stability of this protein or its expression participates in the degenerating cascade. In this context, little is known about the regulation of FAIM-L expression. We have reported that the MEK/ERK pathway is involved in the upregulation of FAIM-L after NGF-induced PC12 differentiation 1 , and, recently, it has been shown that the neural specificity of its expression is governed by NSR100 splicing factor 25 . However, these observations fall short of explaining the regulation of FAIM-L that occurs in the context of axon degeneration. Thus, it would be of interest to further characterize how this protein is regulated during this process in order to shed light on the mechanism that switches FAIM-L-induced XIAP stabilization on and off.
Taken together, our results indicate that, by regulating XIAP protein levels, FAIM-L plays a critical role in regulating non-apoptotic neural events, including axonal degeneration and pruning, and synaptic plasticity events, such as LTD. Moreover, through a mechanism that requires further clarification, we propose that FAIM-L participates in synaptic transmission.
Lentivirus production. Lentiviruses were produced as described by Segura et al. 1

Hippocampal neuron cultures and infection.
Hippocampal neuron cultures were prepared from mouse embryos at embryonic day (E) 15-16. Brains were dissected in PBS containing 0.6% glucose, and hippocampi were dissected out. After trypsin (Invitrogen) and DNase treatment (Roche Diagnostics), tissue pieces were dissociated, and 50,000 cells were seeded onto 12-mm diameter coverslips coated with 0.5 mg/ml poly-L-lysine (Sigma-Aldrich). Neurons were cultured for 15-18 days in Neurobasal medium supplemented with B27 (Life Technologies), glutamine, 20 U/ml penicillin and 20 μ g/ml streptomycin (Sigma-Aldrich, Barcelona, Spain). The medium was further supplemented by adding 1/5 of the volume with conditioned medium from mature (> 14 DIV) hippocampal cultures.
Two consecutive lentiviral infections were performed. XIAP shRNA, FAIM-L shRNA or scrambled infection was performed 3 days before the internalization assay; FAIM-L-EGFP or empty-EGFP lentiviral infection was performed 24 h later. Medium containing viruses was changed 5 h after infection. The size of infected cells did not change under the conditions assayed. The percentage of infected cells (GFP-positive) was around 75%, and the efficiency of both FAIM-L overexpression and XIAP down-regulation was assessed by immunocytochemistry.

Campenot chamber assay of dorsal root ganglia neurons and posterior infection. The
Campenot chamber assay was performed as described 28 , with minor modifications. In brief, poly-L-ornithine and laminin-coated Aclar embedded coverslips (Electron microscopy sciences) were scratched with a pin-rake (Tyler research), and a three-compartment Teflon divider was placed on silicone grease. Dissociated sensory neurons from E13 dorsal root ganglia (DRG) were plated in the central compartment using medium supplemented with 10 ng/ml human NGF. The distal compartments were filled with medium containing 75 ng/ml human NGF. The next day, cultures were treated with 5 μ M cytosine arabinoside. On days 5 and 7, cultures were infected with the corresponding lentiviruses expressing scrambled, XIAP or FAIM-L shRNAs and EMPTY EGFP or FAIM-L EGFP overexpression vectors, respectively. To trigger local axonal degeneration, NGF-containing medium from axonal compartments was replaced with medium containing sheep anti-NGF 1:50 (Abcam). Cultures were fixed with 4% paraformaldehyde 24 h after NGF removal and processed for β III-Tubulin (Covance) (1:4000) and GFP (Life technologies) (1:500) immunofluorescence. Alternatively, cells were collected for Western blot analysis.
Dorsal root ganglia explant culture. Explants culture were obtained from mouse embryos at E13.
Embryos were dissected in L-15 and DRG were dissected out. Explants were plated on 48-or 24-well plates, which had been previously coated with poly-L-ornithine and laminin, for protein and immunochemistry analysis, respectively. After dissection, explants were seeded with medium containing 25 ng/mL NGF and infected with lentiviral vectors expressing EMPTY-EGFP or FAIM-L EGFP. After 8 h, the medium was replaced by medium supplemented with 5 μ M cytosine arabinoside. At DIV 2, explants were subjected to NGF withdrawal, and protein and immunochemistry analysis were performed at 8 and 24 h.

Chem-LTD Induction.
To induce chemical long-term depression (Chem-LTD) in pyramidal neuronal cultures, NMDA was applied at 50 μ M for 5 min onto coverslips in physiological recording solution at room temperature. Upon removal of NMDA-containing solution, coverslips were superfused with physiological solution containing blockers, and recordings were performed at time points ranging from 1 to 20 min. Data were grouped in 5-min clusters (1-5 min, 6-10 min, 11-15 and 16-20 min after NMDA application).
Electrophysiology Data Analysis and Statistics. mEPSCs were filtered at 2 KHz and digitized at 5 KHz.
Data were analyzed using IGOR Pro (Wavemetrics Inc., OR, USA), together with Neuromatic (Jason Rothman, UCL). Events were detected using an amplitude threshold of 4-6 pA set according to the baseline current variance. For mEPSC amplitude analysis, only events with a monotonic fast rise (< 1.2 ms) and uncontaminated decay were included. When analyzing frequency, any event irrespective of rise time or overlapping decays was included. Statistical analysis was performed using GraphPad Prism version 5.0d for Mac OS X (GraphPad Software, San Diego California USA, www.graphpad.com). Data are presented in the text as the mean ± SEM from n recorded cells and in the figures as bar plots of the group mean, with error bars denoting the SEM. Statistical significance between two groups was examined using the non-parametric Mann-Whitney U test.