Great paper here, summarizing the research on why maintaining weight loss is so difficult AND also the foolishness of calorie restriction.
This article discusses research on several factors that may contribute to weight regain following weight loss achieved through behavioural interventions, including adipose cellularity, endocrine function, energy metabolism, neural responsivity, and addiction-like neural mechanisms. All of these mechanisms are engaged prior to weight loss, suggesting that these so called "anti-starvation" mechanisms are activated via reductions in energy intake, rather than depletion of energy stores.
The bolded part is the key part of this paper. The author's are arguing that it is not weight loss per se that sets us up for weight regain, but rather it is the calorie restriction. I posted on another study earlier this year which also seemed to support this idea, here, in that post, we found that mice who lost weight by 40% calorie restriction had a dramatic increase in AgRP, while mice that were allowed to eat ad-lib on a new diet did not increase AgRP but still lost considerable weight. Whats funny is that the author of the above study makes no mention of the AgRP study I blogged about, so unwittingly he is on the right tracks. imo.
The whole paper is golden so if you can get the full text its well worth a complete read. I tried and failed to paraphrase it, every sentence has useful information. So I just copy paste the beginning of the discussion and the conclusion...
Changes in adipose cellularity and addiction-like neural habituation result from chronic overconsumption and appear irreversible via behavioral weight loss [24,34,122,129]. Thus, these factors are not activated to prevent weight loss but serve to encourage preservation of highest sustained body weight, and may actually promote indefinite increases in energy storage. Alterations in endocrine function (e.g., decreases in leptin and increases in ghrelin), decreases in energy expenditure, and increases in neural responsivity to high-calorie food cues all occur within 24 h of caloric restriction (Table 1) [57,84,132].
Regardless of when these mechanisms are activated, each has the potential to exert a [neuro]biological influence that may reduce an obese or formerly obese individual's ability to maintain behavioral weight losses and promote weight regain at least to the individual's highest sustained lifetime weight. These influences also carry the expected weight regain promoting behavioral correlates.
Weight-reduced vs. never-obese subjects report increased food craving , a decreased perception of amount eaten , decreased postprandial satiety  and an increased preference for calorically dense foods .With these additional biological influences encouraging the consumption and storage of energy, it is not surprising that weight regain following behavioral weight loss occurs at a faster rate than initial weight gain [135,137].
These mechanisms appear not to be part of a highly sensitive homeostatic feedback system designed to regulate body weight at any particular “set point,” but mechanisms either acquired via excess weight gain or enacted almost immediately via reduced caloric intake. Importantly, these mechanisms operate irrespective of the adequacy of energy stores. Thus, these mechanisms may be more accurately described as anti-weight loss mechanisms, rather than anti-starvation mechanisms per se.
We have presented evidence that the likelihood of weight regain in weight-suppressed obese and formerly obese individuals may be increased by a confluence of biological mechanisms, including adipose hyperplasia, increased metabolic efficiency, changes in neuroendocrine signaling (e.g., decreased satiety signaling), and changes in neural responsivity to both food cues (e.g., increased reward-related or decreased inhibitory anticipatory responsivity) and food intake (e.g., decreased consummatory reward through habituation to the rewarding aspects of palatable food).
These biological pressures that may undermine weight loss efforts and promote weight regain are almost immediately enacted in obese individuals attempting even modest and healthy weight reduction. Further, these mechanisms operate invariably and appear to defend an individual's highest sustained body weight. Thus, it is the opinion of these authors that these mechanisms would be more accurately described as anti-weight loss mechanisms rather than anti-starvation mechanisms. Regardless, obese individuals face an extreme uphill battle in having to overcome powerful biological drives that appear insurmountable via behavioral interventions, illustrating the critical importance of obesity prevention efforts for normal and overweight individuals. This may be particularly pertinent to parents of overweight children, who are significantly more likely to become obese adults.
p.s. thanks to bill/caloriesproper for the full text.
EDIT -- Added the parts on adipose morphology and leptin
Excess weight gain typically leads to changes in body composition, including significant alterations in adipose cellularity. Although increases in body mass index (BMI) do not directly predict an absolute increase in body fat content , elevated body weight is generally associated with an increase in the diameter of fat cells (adipocyte hypertrophy), as well as greater amounts of fat stored within adipocytes [20,21]. Most literature points to adipocyte hypertrophy as the main feature of obesity; however, alterations in adipocyte number may also be important [22,23]. Upon reaching an upward critical limit in fat cell volume, enlarged adipocytes secrete paracrine factors that induce preadipocyte proliferation (hyperplasia) [24–26]. Thus, excess caloric intake* may lead to increases in fat cell size and subsequent increases in fat cell number [20,26,27].
Recent evidence suggests that this increase in fat cell number may occur in overweight individuals . However, the preponderance of evidence suggests that hyperplasia occurs primarily in clinically severely obese individuals [27,29,30]. Thus, if hyperplasia is associated with weight regain, this effect may be relegated to weight regain following weight loss in [formerly] clinically severely obese individuals, for whom returning to a lean bodyweight through behavioral weight loss is exceedingly difficult .
With behavioral weight loss, adipocyte hypertrophy decreases; however, the hyperplasia remains [20,29,32–35]. Thus, weight loss dieting may reduce the size but not the number of fat cells. A lack of programmed cell death may be responsible for the failure of reductions in fat mass via nonsurgical means to reduce adipocyte number [20,33]. Therefore, relative to never-obese individuals, weight-suppressed [formerly] obese individuals (particularly clinically severely obese individuals) may be left with a significantly greater number of adipocytes, which cannot be reduced via behavioral weight loss . See Table 1.
Liposuction is the only known treatment able to reduce adipocyte number, but carries high complication rates . It is not yet definitively known whether hyperplasia encourages weight regain in weight-suppressed individuals. There is some evidence to suggest that the presence of smaller adipocytes may encourage weight regain by decreasing the overall rate of fat oxidation and increasing the retention of ingested fuel [37–41]. Normally, during times of energy deprivation, fat stores break down triglycerides into their individuals components, glycerol and free fatty acids , which generate energy for the cell. However, the rate of fat breakdown (lipolysis) appears to be related to adipocyte size and cellular surface area ; smaller cells exhibit lower rates of basal lipolysis . Therefore, if size-reduced adipocytes are modified to break down less and store more fat, these cells may expand and promote further proliferation. Although still speculative, there is some evidence to suggest that these cells may be predisposed to reach a particular mean size, allowing them to store similar amounts of fat as previously formed adipocytes [25,34].
However, small adipocyte number may be sufficient to observe a clinically significant effect in only a percentage of obese (i.e., clinically severely obese) individuals.An additional line of evidence reports higher levels of insulin in newly size-reduced adipocytes [44,45]. Insulin, which is excreted from pancreatic beta cells in response to rising levels of glucose in the bloodstream, facilitates a preferential utilization of carbohydrates to meet the cell's energy requirements [40,46–48]. Further, insulin inhibits lipolysis  and stores triglycerides in adipocytes (lipogenesis) .
Interestingly, although insulin sensitivity seems to improve in weight reduced individuals, fat metabolism slows, potentially in an attempt to preserve energy stores [37,38,49,50]. As a result of these changes in carbohydrate and fat utilization, an abnormal accumulation of triglycerides may give rise to a higher net fat cell content and elevations in body weight [37,38,51–53]. Adipocyte size is also correlated with plasma leptin concentrations, which have been shown to affect weight loss maintenance . Relative to control, formerly obese weight-suppressed participants were found to have reduced fat cell volume and serum leptin levels, despite almost identical body fat percentages .
Because smaller adipocytes in formerly obese individual smay be secreting less leptin following behavioral weight loss [28,35,37,55], an association between increased number of smaller adipocytes and leptin insufficiency has been proposed
Although leptin levels are not entirely depleted in weight-suppressed formerly obese individuals, their secretions are much more attenuated relative to lean subjects who undergo caloric restriction [35,55]. Thus, with reductions in leptin secretion, heightened appetite and excess food intake may lead to weight regain [28,54]. The potential role of leptin in weight regain is further discussed below.
Leptin levels are reduced within 24 h of energy restriction  and a number of studies report greater reductions of leptin than would be expected for given losses of adipose tissue [34,35,58]. It has been suggested that leptin's primary role is the prevention of starvation, rather than weight regulation per se, questioning the notion of “leptin resistance” . Reductions in leptin levels appear to trigger a starvation defense response, despite the persistence of abundant fat stores . Evidence suggests that there may be a threshold below which the “anti-starvation” action of leptin is enacted, and this threshold is proposed to increase concurrently with increases in adipose tissue .
Thus, weight loss dieting in obese individuals may lead to leptin depletion to sub-threshold levels, despite the persistence of relatively high levels of leptin. Sub-threshold leptin levels result in reductions in metabolic rate and physical activity , as well as increases in hunger and food intake . Thus, behavioral weight loss and weight loss maintenance are accompanied by physiological attributes that resemble those of a leptin-deficient animal: lower energy expenditure, increased hunger, reduced thyroid metabolism, and diminished sympathetic nervous activity [60,61].
*I think that excess caloric intake quote is very ignorant of the author's. It is much more appropriately stated as excess storage of fat and/or insufficient mobilization of fat.