|Home | About | Journals | Submit | Contact Us | Français|
Both chronic stress and antidepressant medications have been associated with changes in body weight. In the current study, we investigate mechanisms by which stress and antidepressant medications interact to affect meal patterns. A group of mice was subjected to the chronic social defeat stress model of major depression followed by fluoxetine treatment and was subsequently analyzed for food intake using metabolic cages. Here, we report that chronic social defeat stress increases food intake by specifically increasing meal size, an effect that is reversed by fluoxetine treatment. In an attempt to gain mechanistic insight into changes in meal patterning induced by stress and fluoxetine, fasting serum samples were collected every four hours over a 24-hour period and acyl-ghrelin, leptin, and corticosterone levels were measured. Chronic stress induces a peak in acyl-ghrelin levels just prior to lights off, which is shifted in mice treated with fluoxetine. Taken together, these results indicate that stress increases food intake by decreasing satiation, and that fluoxetine can reverse stress-induced changes in meal patterns.
Multiple studies now indicate that certain mental disorders including major depressive disorder (Anderson, Cohen, Naumova, Jacques, & Must, 2007; Beydoun & Wang, 2010; Cizza, 2011; de Wit, et al., 2010; Gavin, Rue, & Takeuchi, 2010; Merikangas, Mendola, Pastor, Reuben, & Cleary, 2011; Papakostas, et al., 2005; Rivera, et al., 2011) and post-traumatic stress disorder (Chwastiak, Rosenheck, & Kazis, 2011; Coughlin, 2011; Coughlin, Kang, & Mahan, 2011; Dedert, et al., 2010; Perkonigg, Owashi, Stein, Kirschbaum, & Wittchen, 2009; Scott, McGee, Wells, & Oakley Browne, 2008; Vieweg, et al., 2006; Vieweg, et al., 2007; Wilson, 2010) significantly increase the risk of developing obesity. In addition, several medications, including antidepressants, used to treat these psychiatric disorders are also associated with changes in body weight (Serretti & Mandelli, 2010). Despite the association between mental illness, psychotropic medications, and weight changes, relatively little is known about the underlying biological mechanisms that mediate these interactions.
Rodent models of major depression have confirmed the importance of chronic stress in body weight homeostasis (Tamashiro, Nguyen, & Sakai, 2005). The visible burrow system, one of the best-studied models of stress, induces changes in food intake and body composition (Tamashiro, Hegeman, et al., 2007; Tamashiro, Hegeman, & Sakai, 2006; Tamashiro, et al., 2004; Tamashiro, Nguyen, et al., 2007). In this model, group-housed rats form a dominance hierarchy. While dominant male rats maintain normal body weight and food intake, subordinate males significantly reduce their food intake and lose weight during the testing period. Interestingly, during the recovery phase following removal from the visible burrow system, subordinate male rats regain their lost weight mostly as adipose tissue and develop obesity and concurrent elevated leptin levels (Tamashiro, Nguyen, et al., 2007). During the stressful hierarchy-formation phase, subordinate males reduce food intake primarily via decreased meal frequency and meal size. In contrast, during the recovery phase, subordinate males increase food intake via increased meal size (Melhorn, et al., 2010).
Using the chronic social defeat stress (CSDS) model in mice, which shares features with both post-traumatic stress disorder and major depression with co-morbid anxiety (Avgustinovich, Gorbach, & Kudryavtseva, 1997; Krishnan, Berton, & Nestler, 2008), our group has analyzed the effect of psychosocial stress on body weight regulation. In this model, a substantial proportion of mice subjected to repeated social aggression develop behavioral deficits similar to symptoms observed in human mental illness, including social avoidance, changes in locomotor activity, impairments in cognition, hyperphagia, decreased preference for natural reward, increased preference for drug reward, and anxiety-like behaviors (Chuang, Cui, et al., 2010; Cruz, Quadros, Hogenelst, Planeta, & Miczek, 2011; Rygula, et al., 2005; Yu, et al., 2011). These so-called “susceptible” mice (Krishnan, et al., 2007) display multiple metabolic abnormalities including body weight gain, redistribution of body fat into internal organs, and insulin and leptin resistance (Chuang, Cui, et al., 2010; Chuang, Krishnan, et al., 2010). The circulating hormone ghrelin appears to be an important mediator of certain CSDS-induced behavioral and metabolic disturbances. CSDS increases acyl-ghrelin levels, and mice lacking ghrelin receptors exhibit exaggerated social avoidance behavior after CSDS (Lutter, Sakata, et al., 2008). More recently, we have shown that following CSDS, wild-type mice, but not ghrelin receptor-null littermates, display conditioned place preference for high fat diet. Importantly, restoration of ghrelin receptor expression specifically in catecholaminergic neurons reduces social isolation, but at the expense of also restoring conditioned place preference for high fat diet (Chuang, et al., 2011). These findings suggest that ghrelin may be an important mediator linking changes in appetite to mood symptoms.
Because CSDS increases food intake and conditioned place preference for high fat diet in susceptible mice (Chuang, Cui, et al., 2010; Chuang, et al., 2011), it appears to model aspects of depression with atypical features. Therefore, we wanted to determine if antidepressant treatment could reverse the effects of CSDS on food intake in this subpopulation of animals. We chose fluoxetine because it is a widely prescribed antidepressant and because several initial studies reported that fluoxetine suppresses food intake in rodents (Gamaro, Prediger, Lopes, Bassani, & Dalmaz, 2008; Grignaschi & Samanin, 1992; Lightowler, et al., 1996; Wong, Reid, & Threlkeld, 1988), and human studies also suggest that fluoxetine acutely suppresses food intake (Serretti & Mandelli, 2010). For example, during a one-year maintenance trial, fluoxetine treatment was associated with a modest initial weight loss during the first four weeks of treatment (Michelson, et al., 1999).
In the current study, we tested the hypothesis that fluoxetine treatment will normalize food intake in mice subjected to CSDS. We exposed mice to CSDS or control handling conditions. Mice displaying social avoidance were randomized to receive either fluoxetine or placebo. Multiple behavioral and metabolic parameters were then measured, including meal patterning and circulating levels of serum hormones known to regulate food intake and body weight homeostasis.
Animals were housed in the UT Southwestern vivarium in a temperature-controlled environment on a 12 h light/dark cycle (lights on 0600) and maintained on regular chow (4% fat diet #7001, Harlan-Teklad, Madison, WI).
All animal procedures were approved and carried out in accordance with both the UT Southwestern Institutional Animal Care and Use Committee (IACUC) guidelines and “Principles of laboratory animal care” (NIH publication No. 86-23, revised 1985)
The CSDS paradigm used in the current study was recently reported (Lutter, Krishnan, et al., 2008). Eight-week-old male C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) were group-housed and acclimated to the vivarium 1 week prior to the commencement of social defeat. Mice were exposed to a different CD1 strain aggressor mouse (Charles River Laboratories, Wilmington, MA) each day prior to lights out (1800) for 5 min over a total of 10 days. After 5 min of physical contact, test mice were separated from the aggressor and placed across a plastic separator with holes, where they remained in sensory contact with the CD1 aggressor for the remainder of the 24 hours, but physical contact was prevented. Control mice were handled daily in the palm of the hand for 30 seconds and housed in equivalent cages with members of the same strain. After the last defeat, all mice were housed individually and a social interaction task was performed to measure the behavioral consequences of chronic defeat stress.
The social interaction task involved placing mice in a new arena (44 cm × 44 cm) with a small animal cage (5 cm × 10 cm) at one end. Their movements were tracked for 2.5 min in the absence of another mouse, followed by 2.5 min in the presence of a caged, unfamiliar CD1 mouse. Social interaction was quantified by comparing the amount of time the test mouse spent in the interaction zone near the small animal cage in the presence vs. in the absence of the target CD1 mouse, as determined using Ethovision 3.0 software (Noldus, Leesburg, VA). Thirty-two out of 120 mice subjected to CSDS did not display social avoidance (defined as an interaction ratio < 100%) and were removed from further study. Five out of 120 control mice displayed social avoidance and were removed from the study. Twenty-four hours following social interaction testing, the remaining mice were implanted with either a fluoxetine or placebo pellet in the dorsal interscapular region under brief isoflurane anesthesia. Custom-made continuous release pellets (Innovative Research of America, Sarasota, FL) were used to deliver a chronic dose of fluoxetine (15 mg/kg/day) for 20 days. Previous studies showed that pellets prepared in this manner produce physiologically relevant concentrations of fluoxetine 20 days after implantation (Covington, et al., 2009).
Experiments were conducted in the UT Southwestern Metabolic Phenotyping Core using metabolic cages (TSE Systems, Chesterfield, MO). A representative cohort of mice from all four groups of the experimental population was pair matched for body weight and lean body mass (Supplemental Figure 1B–C) to correct for presumed body composition-related differences in food intake. These mice were transferred to the metabolic phenotyping core for 4 days of acclimation and then placed in metabolic cages for 4 days with access to regular chow. Meal analysis over a 48-hour period was assessed using Lab Master Software (TSE Systems, Chesterfield, MO).
Blood samples were collected across the circadian cycle every four hours starting one hour after lights on. Animals from each time point were fasted four hours prior to sacrifice by cervical dislocation in order to control for the effects of food intake on leptin and ghrelin levels. Because of the relatively large amount of serum needed to measure ghrelin levels, we could only obtain the amount of blood required for hormone analysis by collecting blood samples as a terminal procedure and not through serial blood draws. Trunk blood was immediately collected in EDTA-coated tubes (BD Biosciences, San Jose, CA) and was further processed by the addition of 4-(hydroxymercuri)benzoic acid sodium salt (#55540, Sigma Aldrich, St. Louis, MO) to preserve neuropeptide hormones. A fraction of blood sample was set aside and acidified by the addition of 0.1 N HCl in order to stabilize acyl-ghrelin. Serum levels of leptin and corticosterone were measured as previously described (Chuang, Cui, et al., 2010; Chuang, Krishnan, et al., 2010), with intra- and inter-assay variabilities of 5% and 12% for leptin, and 3% and 5% for corticosterone, respectively. Levels of acyl-ghrelin were quantified using a commercially available enzyme immunoassay kit (#10006307, Cayman Chemical, Ann Arbor, MI), with an intra- and inter-assay variability of 4% and 9%, respectively. All hormone levels were analyzed in duplicate.
Data are reported as mean±SEM. Statistical analyses were performed using Student’s t-test, repeated measures analysis of variance (RMANOVA) or two way ANOVA followed by Bonferroni post hoc tests. All statistical analyses were performed using Prism (v 5.0, GraphPad Software Inc., San Diego, CA) software. Statistical significance was defined as p<0.05.
Eight-week-old male C57BL/6J mice were subjected to either 10 days of CSDS or control handling conditions (Figure 1A). After this 10-day period, defeated mice that displayed social avoidance (“CSDS” mice, Supplemental Figure 1A) and nonstressed controls were randomized into two equal groups that received a subcutaneous pellet releasing a 20-day supply of either fluoxetine or placebo. This protocol resulted in four groups of mice with a large number: control-placebo (n = 41), control-fluoxetine (n = 40), CSDS-placebo (n = 39), and CSDS-fluoxetine (n = 37). As previously reported (Berton, et al., 2006), chronic fluoxetine treatment reversed social avoidance 29 days after the last bout of social stress (Supplemental Figure 1A). We then sought to determine whether chronic social stress affects meal patterning, and if fluoxetine influences food intake in these mice. However, there has been much uncertainty on how to interpret measures of food intake when differences in body composition exist (Butler & Kozak, 2010; Tschop, et al., 2011). Therefore, to rigorously answer this question, a subset of mice from all four groups was pair matched for body weight and lean body mass (Supplemental Figure 1B–C) to correct for presumed body composition-related differences in food intake, and food intake was measured over a 4-day period using metabolic cages.
As shown in Figure 1, there were no differences noted in total chow consumed, meal number, meal size, or satiety ratio among groups during the light phase when mice are typically inactive. In contrast, mice exposed to CSDS plus placebo demonstrated an increase in total chow consumed during the dark phase (Fig. 1B). Meal pattern analysis revealed that this increase was the result of fewer meals (Fig. 1C) and increased meal size (Fig. 1D). The satiety ratio (g of food per meal/intermeal interval) was the same in all groups indicating that CSDS increases food intake primarily by impairing satiation (the process of intrameal termination) during the dark phase without affecting satiety (the length of appetite suppression after a meal). The effects of fluoxetine on meal patterning were restricted to the CSDS-exposed mice, for which fluoxetine normalized CSDS-induced changes in total chow consumed, meal number, and meal size (Fig. 1B–D).
The metabolic cage data indicate that CSDS increases food intake in susceptible mice, and that fluoxetine preferentially reduces hyperphagia in these mice. Furthermore, CSDS appears to increase food intake only during the dark phase, suggesting a diurnal effect. For further mechanistic insight into the effects of CSDS and fluoxetine on meal patterning, we next analyzed levels of circulating hormones after CSDS. Acyl-ghrelin, leptin, and corticosterone levels have previously been demonstrated to be disturbed in susceptible mice following CSDS (Chuang, Cui, et al., 2010; Krishnan, et al., 2007; Lutter, Sakata, et al., 2008). However, the measurements in these previous studies were made at a single time point, and thus likely were not reflective of time-of-day differences in appetitive behavior, as detected in the current study. Because the current study was designed to collect samples over a 24-hour period, we were therefore able to measure serum hormone levels across the circadian cycle.
Consistent with our previous observations (Lutter, Sakata, et al., 2008), acyl-ghrelin levels are increased in susceptible mice exposed to CSDS in the late light phase just prior to lights off (Fig. 2A, ZT9), with trends towards increased levels in this group during the preceding two timepoints. Fluoxetine treatment normalizes acyl-ghrelin levels after CSDS at ZT9. A significant peak of acyl-ghrelin expression at ZT17 was noted only in the CSDS-fluoxetine group (the functional significance of this peak will be addressed in the Discussion). We next analyzed leptin levels, previously shown to be reduced after CSDS (Chuang, Cui, et al., 2010), and found significant reductions in leptin in both CSDS groups at ZT5, ZT9, and ZT13. Fluoxetine treatment did not significantly alter leptin levels in either control or CSDS groups (Fig. 2B). Chronic social stress reduced morning levels of corticosterone in both CSDS groups (Fig. 2C), consistent with previous observations (Krishnan, et al., 2007). Fluoxetine also caused a significant reduction of corticosterone levels at ZT21 in both control and CSDS groups. No other effects of CSDS or fluoxetine were noted in corticosterone levels at other time points.
Though the link between stress and body weight change has long been appreciated, less is known about the neurobiological mechanisms that underlie the effect of antidepressant treatment on body weight changes. We hypothesized that fluoxetine treatment would reverse the hyperphagia observed in mice following CSDS. In the current study, we report several novel observations on the association of stress, antidepressants, and food intake among stress-susceptible animals.
Previously we reported that mice displaying depression-like behavior after CSDS exhibit pronounced hyperphagia (Chuang, Cui, et al., 2010; Chuang, Krishnan, et al., 2010). Here we find that CSDS increases food intake primarily by increasing meal size, suggesting a specific deficit in satiation or the processes that terminates a meal. This finding is consistent with previous work in the visible burrow system in which subordinate male rats increase food intake during the recovery phase by increasing meal size (Melhorn, et al., 2010). In contrast, other stress models lacking a social component, such as the chronic unpredictable stress and chronic restraint stress models, report a decrease in food intake in response to stress (Chiba, et al., 2012; Varga, et al., 2011). It may be the case that psychosocial stress, rather than more neutral environmental stressors, uniquely contributes to the development of hyperphagia. For this reason, our key finding of enhanced meal size has several potential clinical implications. First, it indicates that behavioral techniques aimed at improving satiation and not satiety (how long one’s appetite remains suppressed after a meal) should be the focus of treatment of obesity in patients with mental disorders (Blundell & MacDiarmid, 1997; Green, Delargy, Joanes, & Blundell, 1997). Secondly, it begins to delineate the neural circuits that control different aspects of food intake following exposure to stress. Ghrelin signaling directly on catecholaminergic neurons affects mood symptoms and conditioned place preference for high fat diet following CSDS (Chuang, et al., 2011). In contrast, changes in satiation may be regulated by the coordinated action of the vagus nerve and gut hormones, which serve as peripheral sensors of energy intake to control meal size (Berthoud, 2008).
Our observation that meal size is restored to control levels in CSDS animals treated with chronic fluoxetine carries several implications for the role of serotonin in satiation mechanisms. Serotonin has been demonstrated to be an important regulator of feeding in rodents. Administration of serotonin receptor agonists activates pro-opiomelanocortin (POMC)-expressing and inhibits agouti-related peptide/neuropeptide Y (AgRP/NPY)-expressing neurons in the arcuate nucleus of the hypothalamus, resulting in a suppression of food intake (Heisler, et al., 2006; Lam, et al., 2008). Additionally, animals lacking melanocortin 4 receptors (MC4R), as well as transgenic mice overexpressing the MC4R antagonist agouti, are unresponsive to the anorectic effects of augmented serotonin signaling (Heisler, et al., 2006; Lam, et al., 2008), which suggests an essential role for melanocortin signaling in the regulation of food intake by serotonin. These studies provide a general framework through which to interpret our finding that fluoxetine suppresses meal size; however, it is unclear why we do not observe a similar reduction among control mice. In nonstressed rats, it was also reported that daily food intake in the terminal phase of a chronic fluoxetine regimen was unaltered by the drug (Gamaro, et al., 2008). The mechanism explaining this discrepancy between control and CSDS animals may involve important neural adaptations to repeated social stress that sensitize feeding circuits to the appetite-suppressing effects of fluoxetine. This is a plausible hypothesis in the context of CSDS, which is known to induce long-lasting disturbances in limbic circuitry responsible for orchestrating both emotional and appetitive behaviors (Berton, et al., 2006; Tsankova, et al., 2006). As an example, various models of both acute and chronic stress, including CSDS, result in increased activity of mesocorticolimbic dopaminergic circuitry (Krishnan, et al., 2007; Lammel, Ion, Roeper, & Malenka, 2011; Valenti, Lodge, & Grace, 2011), a circuit which is thought to coordinate motivated behaviors associated with feeding (Palmiter, 2007). Moreover, in the CSDS model, hyperactivity of ventral tegmental area (VTA) dopaminergic neurons is restored to control, nonstressed levels by a chronic fluoxetine regimen similar to the one we utilize here (Cao, et al., 2010). Of note is the fact that dopamine neuron firing among nonstressed animals is unaffected by fluoxetine (Cao, et al., 2010), again pointing to a phenotype-specific effect of this antidepressant. Interestingly, ghrelin receptors expressed on presynaptic dopaminergic neurons of the VTA are known to provide excitatory input to this neuronal population (Abizaid, et al., 2006), and infusion of ghrelin directly into the VTA results in increased food intake (Naleid, Grace, Cummings, & Levine, 2005). It is possible that, in the CSDS model, fluoxetine normalizes the stress-induced decrease in satiation by reducing acyl-ghrelin levels, which in turn function to reverse the neuronal adaptations associated with CSDS. While this possibility has yet to be fully examined, we speculate that CSDS alters the basic function of satiation circuits in such a way as to render them plastic to modulation by fluoxetine.
Ghrelin and leptin levels have previously been shown to be increased and decreased, respectively, in rodent models of stress including chronic unpredictable stress and CSDS (Asakawa, et al., 2001; Chuang, Cui, et al., 2010; Chuang, et al., 2011; Garza, Guo, Zhang, & Lu, 2011; Kristenssson, et al., 2006; Lu, Kim, Frazer, & Zhang, 2006; Lutter, Sakata, et al., 2008; Ochi, et al., 2008; Patterson, Ducharme, Anisman, & Abizaid, 2010; Rouach, et al., 2007). Here, our findings indicate that acyl-ghrelin, but not leptin, levels are altered after fluoxetine exposure following CSDS (Fig. 2A and 2B). It should be noted that serum hormone levels were measured from the overall pool of mice and not the sub-group selected for metabolic cage monitoring, although similar levels of hyperphagia were observed in both groups, suggesting that the groups are likely comparable.
Fluoxetine treatment after CSDS is associated with reduced ghrelin levels at ZT 9, but with the appearance of a peak at ZT 17. One possible mechanism for this action was reported by the Inui Laboratory, who demonstrated that selective serotonin reuptake inhibitors, including fluoxetine, decrease acyl-ghrelin levels via central activation of the 5-HT2c receptor (Fujitsuka, et al., 2009). Our data suggest that 5-HT2c receptor signaling may be important for the timing of the acyl-ghrelin peak and that fluoxetine treatment shifts the peak until late in the dark phase (the potential implications of this shift on feeding behaviors will be discussed below). We have previously shown that ghrelin receptor signaling is required for the increased food intake and conditioned place preference for high fat diet induced by CSDS (Chuang, et al., 2011; Lutter, Sakata, et al., 2008). Further studies will be necessary to determine if fluoxetine normalizes satiation after CSDS via changes in ghrelin signaling or through a distinct mechanism.
We also show that changes in leptin and acyl-ghrelin levels only occur at specific times of the day. After CSDS, leptin levels are significantly lower than in control mice during the late light/early dark phase. Meanwhile, CSDS induces a peak in ghrelin levels in late light phase at ZT9. This well-timed increase in the relative ratio of ghrelin/leptin may explain why feeding differences are only observed in the dark phase. Recent work from the Sternson Laboratory identified a ‘flip-flop’ memory circuit in AgRP/NPY neurons in the arcuate nucleus of the hypothalamus (Yang, Atasoy, Su, & Sternson, 2011). In their model, acyl-ghrelin acts on pre-synaptic neurons to increase glutamate release onto AgRP/NPY neurons in a feed-forward loop. This excitatory neurotransmission increases feeding responses for hours, even after withdrawal of ghrelin, until leptin terminates feeding by acting on POMC-expressing neurons to inhibit AgRP/NPY neuronal activity via opioid release. Our data suggest that CSDS reduces leptin levels from ZT5 to ZT17, creating a 12-hour ‘window’ in which hyperphagia can occur. However, feeding does not happen until late in the light phase when acyl-ghrelin levels peak. The feeding bout would then continue through the early dark phase until leptin levels rise at ZT17 in the late dark phase. This model would also explain why a shift in the peak of acyl-ghrelin levels in the CSDS-fluoxetine group from ZT9 to ZT17 is not associated with increased food intake because leptin levels have returned to baseline and no longer permit activation of AgRP/NPY neurons.
One limitation of this study is that our findings only apply to the specific sub-groups that were selected for analysis. Specifically, we included mice susceptible to social avoidance after exposure to CSDS in addition to mice that displayed reversal of social avoidance by treatment with the antidepressant fluoxetine. This decision was made to achieve greater relevance for patients who present in clinical situations. Typically patients presenting for treatment have experienced mood-related symptoms induced by periods of stress and frequently receive antidepressant medications. Therefore, our treatment groups most accurately reflect actual clinical scenarios. Further studies that analyze metabolic parameters in mice ‘unsusceptible’ to the effects of CSDS as well as avoidant mice that do not respond to antidepressant treatment may be warranted to determine the metabolic consequences of these states.
Together these observations support a model in which stress pathways reduce satiation to increase meal size. Additionally, antidepressant treatment may normalize food intake in tandem with mood symptoms among subjects exposed to psychosocial stress. These insights may yield more effective treatments for stress-induced obesity.
A. Social interaction results of animals in study population. (B–F) Additional parameters of mice monitored in metabolic cages. B. Body weight (no significant difference), C. Lean body mass (no significant effect), D. adipose tissue (significant effect of defeat, F1,18 = 6.31), E. Locomotor activity during light phase (significant effect of defeat, F1,18 = 6.19 and significant effect of fluoxetine, F1,18 = 7.12), and F. Locomotor activity during dark phase (significant effect of defeat, F1,18 = 5.41). N = 5–6/group. *p< 0.05, **p < 0.01, ***p < 0.001. Data presented as mean ± S.E.M.
Support. National Institutes of Health: K08 MH084058-1A1 (ML), R01 MH082876 (CAM), R01 MH085298 (JMZ), UL1RR024923, RL1 DK081182, RL1 DK081185, T32-MH076690-05 (EN), and the Medical Scientist Training Program (JK). National Alliance for Research in Schizophrenia and Depression (NARSAD): Young Investigator Award (ML). The University of Texas Southwestern Medical Center: The Disease Oriented Clinical Scholars Program (ML) and the Summer Undergraduate Research Fellowship Program (AK).
We thank Edgaro Falcón, Sadé Spencer, Angela Walker, and Michelle Sidor for technical assistance.
Conflict of Interest: None (JK, J-CC, ESN, AK, AGG, SM, JMZ, ML). CAM has received honoraria from GlaxoSmithKline, Pfizer, and Servier, consulting fees from Orphagen Pharmaceuticals, and research funding from GlaxoSmithKline and Pfizer for unrelated projects.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.