Mouse models of uremia are important tools for in vivo translational studies such as examining the impact of specific genes in transgenic or knockout mice and to validate potential therapeutic interventions prior to pre-clinical testing in humans. The limitations of existing surgical models of uremia in mice, including requirements of surgical skills, demand of animal facilities to support post-operative care, high mortality rate and less flexibility in terms of dynamic alterations in urea levels, prompted us to develop a non-surgical model of renal dysfunction based on dietary intake of adenine. This approach has successfully been used in rats but not in mice because of their reluctance to consume adenine. Accordingly, addition of adenine to a standard mouse chow elicits high morbidity and mortality due to starvation and malnutrition rather than renal failure. In our model, this was circumvented by mixing the adenine in a casein-based chow, in which the casein effectively removed the inherent smell and taste of adenine.
Since the sensitivity to adenine as well as food intake may vary between various strains of mice, we propose a protocol with ad libitum changes in adenine concentration between 0.15-0.20% during the maintenance phase. The accepted variability in blood urea level during the maintenance phase may cause inter-individual variations in kidney function that could affect the phenotype. However such variability is inevitable and present in other uremic animal models as well. On the contrary, providing a target urea interval provides the opportunity to adjust blood urea levels according to the desired outcome or mechanisms of interest.
The apparent improvement in kidney function after lowering the adenine concentration, as indicated by a decline in blood urea and PTH levels, suggests at least a partial reversibility of renal impairment. However we did not examine whether adenine discontinuation translated into long-term histological improvements. Given the severity of the renal histopathological lesions observed after eight weeks of adenine exposure, we anticipate that long-term adenine challenge will cause chronic renal failure with less reversibility as observed in adenine-induced uremia in rats . Regardless, the possibility to short-term modulate kidney function provides significant benefits in terms of the possibility to extend study protocols and to analyze outcome variables in various strata of kidney function. Additional advantages of our model include zero mortality, which limits the number of animals needed for induction, and the small inter-individual variation in renal function that contrasts the 5/6 nephrectomy model [10, 17]. It also provides a good option for researchers with limited surgical competence and/or restrictions in post-operative care.
Some more common non-surgical options to study uremia including radiation nephropathy and administration of nephrotoxic drugs such as folic acid , cyclosporin A  and cisplatin  should be mentioned. However, these models are non-reversible, strain-dependent and of limited use due to systemic toxicity. Genetic mouse models mimicking various aspects of kidney failure are also available, but these are compromised by the need for breeding to create combined backgrounds with other genetically altered mouse strains .
Several different uremic models have also been developed in rats. Advantages with rat models are that collected blood and urine volumes are significantly greater, blood samples at intermediate time points are more easily obtainable and certain organs such as parathyroid glands are readily identified. Another apparent difference is that rats tolerate higher adenine exposure, which reportedly produce 3–5 times higher blood urea concentrations than our model. This may also impact renal histology since rats in additional to tubular damage generally also suffer from extensive glomerular damage which was not found in our model .
There are several biochemical findings of interest in our model. Serum calcium level was unaltered in the adenine-treated mice likely due to a compensatory rise in PTH, which largely mimics the situation in patients with CKD stage 3–4. Normal serum calcium concentrations have also been reported in other CKD models . Another striking finding is the continuous and marked rise in FGF23 although other bone-mineral markers remained more constant during the maintenance phase. This supports the presence of renal derived factors that regulates FGF23 synthesis in bone. Alternatively, the tubular damage may severely hamper intact FGF23 degradation leading to accumulation of circulating FGF23 protein. The pattern of bone markers suggesting an increased bone formation but decreased bone resorption is somewhat unexpected in a uremic model of secondary hyperparathyroidism and anticipated high bone turnover rate. The mechanism(s) underlying reduced bone resorption are unclear but could speculatively be due to impaired osteoclast function as a result of the exceptionally elevated levels of FGF23.
Some potential limitations should be mentioned. We cannot exclude the possibility of systemic toxicity or organ-specific damages caused by the adenine. Because tubular toxicity of adenine metabolites is the underlying mechanism of adenine-induced renal failure [23, 24], our model primary reflects a tubulointerstitial disease whereas the most common cause of CKD in human is glomerular scarring secondary to vascular damage. Thus, our model should not be regarded as a model of CKD per se but rather as a complementary model of renal failure. We did not determine the dietary (caloric) intake, yet based on previous experiments in rats the adenine-fed animals may have somewhat lower overall dietary intake. Further, we did not perform dynamic bone histomorphometry although adenine models in rats produce a high turnover phenotype. Finally, possible strain differences as found in other mouse models  warrant further investigation.