The endocannabinoid system regulates appetite and metabolism [8]. Previous work in rodents suggests an antagonistic relationship between the central endocannabinoid system and leptin [9]. Indeed, administration of leptin to normal Sprague-Dawley rats reduced hypothalamic levels of both AEA and 2-AG [9]. Obese Zucker (fa/fa) rats with defective leptin signaling possess higher hypothalamic 2-AG levels than lean controls, whereas AEA levels are not changed. Elevated hypothalamic 2-AG levels are observed in ob/ob mice (where there is no leptin production) compared to controls. Furthermore, fasting elevates while feeding reduces hypothalamic 2-AG levels in rats [16] and fasting results in reduced leptin levels in the CSF of humans [15, 17]. Based upon these studies, fasting in humans would be expected to produce an inverse gradient between CSF 2-AG and leptin levels. Indeed, such a negative correlation between 2-AG and leptin was observed in the current study, suggesting that the control of 2-AG production by leptin seen in the rodent hypothalamus [9] may also occur in humans.

Leptin is an adipokine that regulates appetite, metabolism, and inflammation and one of its primary roles is to reduce food intake [42–44]. Obesity is associated with chronically elevated leptin levels that may give rise to leptin resistance [45]. Recent work has shown that peripheral endocannabinoid signaling contributes to leptin resistance [11]. In contrast, lack of hypothalamic endocannabinoid signaling leads to leptin resistance in animals fed a high fat diet, whereas its actions are potentiated in chow-fed animals [46]. In our study, the negative correlation between CSF leptin and 2-AG suggests that the reciprocal regulation of central leptin and endocannabinoid signaling remains functionally coupled in overweight fasting humans. Because our serum and CSF samples were obtained during the morning hours, it is tempting to speculate that the inverse relationship between leptin and 2-AG may reflect an increase in the morning orexigenic drive. This is supported by a very recent clinical study demonstrating that 2-AG levels increase while leptin levels decrease during sleep and throughout the morning [47].

Our results are in contrast to the study of Jumpertz et al [28], who did not observe a correlation between leptin and 2-AG levels in the CSF of healthy, middle aged overweight individuals, which the authors attributed to possible leptin resistance in this cohort. Although our sample comprised older patients with osteoarthritis, it is unlikely that the endocannabinoid and leptin levels were affected by the underlying osteoarthritis (see below). However, it is prudent to consider the present findings in context of this co-morbidity and again to suggest that further validation studies are needed. Lastly, the negative correlation between CSF 2-AG and leptin in the fasting subjects would infer that sufficient 2-AG levels derived from the hypothalamus are reflected in the CSF for the effect to be specific to the hypothalamic endocannabinoid signaling process.

Importantly, the CSF endocannabinoid levels reported herein are in line with the literature (Table 3). Most of the previously reported data are for AEA, with mean values for controls ranging from below the limit of detection to 11.65±7.53 pmol/ml (Table 3). Similarly, 2-AG levels ranging from below the limit of detection to 160±110 pmol/mg (61±42 ng/ml) have been reported; and our values are within this range. Lastly, the CSF leptin levels reported herein are similar to those previously observed in subjects without underlying osteoarthritis [48, 49], arguing against osteoarthritis-induced alterations in leptin levels.

Study	Control/Patients	n	AEA	2-AG
Giuffrida et al. [50]	Controls	81	0.007±0.002 b
	Schizophrenic patients: Antipsychotic-naïve	24	0.057±0.011 b
	1 °G antipsychotic treated	36	0.031±0.012 b
	2 °G antipsychotic treated	31	0.062±0.016 b
Pisani et al [51]	Controls	14	5.23±1.81 a
	Untreated PD patients	16	10.7±3.2 a
Sarchielli et al. [52]	Controls	20	0.39±0.09 b	Below detection limit (0.2 pmol/sample) for all groups
	Chronic migraine	15	0.21±0.06 b
	Probable migraine + PAOH	15	0.22±0.05 b
Centonze et al. [53]	Controls	11	2.4±1.3	~75–100 c
	Relapsing-remitting MS	11	20.3±15.7
Leweke et al. [54]	Controls	81	range 0 - ~0.07 c
	Schizophrenics	44	range 0 - ~0.33 c
Di Filippo et al. [55]	Controls	20	0.085±0.0019 a	1.068±0.123 a
	Stable remitting-relapsing MS	20	0.0048±0.0009 a	0.824±0.104 a
	MS during relapse	15	0.0068±0.0009 a	0.997±0.127 a
	Secondary phase MS	15	0.0036±0.0006 a	0.762±0.098 a
Koethe et al. [56]	Controls	81	Median <0.001, IQR <0.001–0.005
	Prodromal psychotic patients	27	Median 0.006, IQR <0.001–0.073
Koppel et al. [57]	Elderly controls and Late onset AD patients	35	Below detection limit	For all cases mean 0.205, range 0.027–0.704
Koethe et al. [58]	Controls before and after sleep deprivation	20	Very low levels shown in Fig 1 of paper
Romigi et al. [59]	Controls	9,6 d	11.65±7.53 a	160±110 a
	Untreated epilepsy	9,6 d	2.55±1.78 a	210±147 a
Pisani et al. [60]	Controls	37	4.59±1.65 a
	De novo PD	38	9.76±3.26 a
	PD, treatment withdrawal	8	11.19±3.23 a
	PD, under treatment	10	6.76±3.41 a
Jumpertz et al. [28]	Controls (BMI 34±8 a kg/m2)	27	0.06±0.08 a	14.2±6.85 a
Morgan et al. [61]	Controls	13	~0.13 c	~10 c
	Light cannabis users	10	~0.16 c	~18 c
	Heavy cannabis users	10	~0.10 c	~23 c

One observation at odds with the literature, however, is the finding of a significant correlation between serum and CSF AEA (but not 2-AG) levels. In four separate studies [28, 50, 56, 61] no such correlation was found, although a positive correlation was seen for schizophrenic patients treated with first-generation antipsychotics [50]. Very recently completed data (i.e. after this manuscript was submitted) by one of us suggests that there is a large inter-sample variation in plasma AEA and 2-AG levels for any given individual when the samples are either taken at a one hour interval or at the same time of day on two different occasions (L. Lindgren, M. Gouveia-Figueira, M.L. Nodding and Christopher J. Fowler, manuscript in preparation). This may account for the differences in plasma:CSF correlation coefficients between the studies. Additionally, it may be possible that in the patient population examined here, the underlying osteoarthritis has produced an effect upon AEA turnover centrally and/or peripherally, thereby inducing a correlation. There is evidence that both inflammatory and anti-inflammatory cytokines can affect components of the endocannabinoid system, such as cannabinoid receptors and fatty acid amide hydrolase [62–64] so in theory this is possible although unlikely because the serum endocannabinoid levels obtained here are similar to those reported for healthy patients without underlying osteoarthritis [65–67]. Furthermore, brain and plasma AEA levels were found not to be altered in an animal model of osteoarthritis [68].

A second potential factor to consider may be the body weight range of the patient sample. In our dataset, only 2 individuals were classified as normal weight (BMI<25) whilst 9 were overweight (25<BMI<30), 22 were obese (30<BMI<40) and 2 were morbidly obese BMI>40). Thus, our findings relate to an overweight/obese population. This may be of importance for the observed and interpreted results, although the literature is somewhat divergent with respect to the association between circulating endocannabinoid levels and body weight / BMI. Indeed, some studies have reported elevated levels of both AEA and 2-AG in overweight and obese individuals [18, 22], some have reported elevations in only one of the endocannabinoids [19, 21, 28, 31], while others have reported no change in either metabolite [24]. The only study investigating the CSF was that of Jumpertz et al. [28], who reported a negative correlation between BMI and AEA, but not 2-AG, levels; but this association was lost in a multivariate analysis adjusting for plasma AEA concentrations. However, our data reveal that central and circulating endocannabinoid levels do not correlate with BMI at least in an osteoarthritis patient population.

In conclusion, the present study investigated the relationship between obesity factors and peripheral and central endocannabinoids in a population of overweight to obese individuals with severe osteoarthritis. Because endocannabinoid levels are dynamically regulated by feeding and fasting in rodents and humans [16, 20, 69], it is clear that the relationship between circulating and central endocannabinoids and obesity-related parameters is complex and most likely dependent upon the patient populations selected. Therefore, it should be noted that the conclusions of our study are restricted to this patient category and some of the differences between the present study and other published data for other patient categories may reflect this fact. Nonetheless, the present study provides evidence for the tight coupling between central endocannabinoid and leptin levels and suggests that the regulation of 2-AG levels by leptin observed in the rodent hypothalamus may likewise exist in humans.