Kimberlites are the deepest mantle magmas to reach the surface of the Earth and the hosts of the major primary diamond deposits. The enigmatic nature of kimberlites owning to poorly constrained triggers of kimberlite magmatism, melt composition, and crystallization conditions, are mostly limited by unknown content and composition of volatiles. The content and ratio of H2O and CO2 are important parameters of magma emplacement. They are essential for the mantle melting and rapid ascent of kimberlites. Volatiles also greatly affect diamond preservation in kimberlites during the ascent to the surface. Apatite has been used as an indicator of volatiles in different magmatic systems. Partitioning of trace elements between apatite and a melt is also sensitive to the carbonate component in the melt. This makes apatite potentially very useful indicator of evolution of melt composition and fluid during kimberlite ascent and emplacement. However, trace element partition coefficients between apatite and melts have been determined for compositions much more SiO2-rich than kimberlites, whereas coefficients in carbonate melts show large discrepancies. This study presents experimentally determined trace element partition coefficients (D) between apatite and synthetic analogues of kimberlite (SiO2 content ranging from 11 to 23 wt%) and carbonatite melts (SiO2 content <1 wt% and 4.5 wt%). The experiments were conducted in piston-cylinder apparatus at 1150o, 1250o, 1350o, at 1 and 2 GPa, and at oxygen fugacity -3.97, 0.27, 4.83 log units relative to fayalite-magnetite-quartz (FMQ) buffer. We demonstrate the increase of DREE for Rare Earth Elements (REE) with increase in SiO2 of the melt and the effect of apatite composition (natural Durango apatite vs. synthetic apatite) on D and substitution mechanism. We found no effect of temperature, pressure, water content, and oxygen fugacity on D in these low SiO2 melts. This allows us to use trace element content of natural apatite to examine the composition of kimberlite magma. We apply the results to apatite from two kimberlite classes: Class 1 composite kimberlite pipe from Orapa cluster (Botswana) with hypabyssal coherent and volcaniclastic kimberlite facies and Class 3 kimberlite pipes from Ekati Mine (Northwest Territories, Canada) filled with effusive coherent kimberlites in Leslie pipe and both volcaniclastic and coherent kimberlites in Boa pipe, and discuss how changes in SiO2 content of kimberlite melt affect apatite saturation. We further apply our experimentally determined DREE to model the composition of evolved kimberlite melt using published bulk rock and apatite compositions in kimberlites.