There is a huge gap between properties of red-sequence selected massive galaxy clusters at z<1 and Lyman-break selected proto-clusters at z>3. It is important to understand when and how the z>3 proto-clusters evolve into passive clusters at z<1. We aim to fill this cluster desert by using the space-based N4(4um) imaging with the AKARI. The z-N4 color is a powerful separator of cluster galaxies at z>1, taking advantage of the 4000A break and the 1.6um bump. We carefully selected 16 promising cluster candidates at 0.9<z<1.7, which all show obvious over-density of galaxies and a prominent red-sequence. At this redshift range, the mid-infrared S15um/S9um flux ratio is an extinction-free indicator of galaxy star formation activity due to the redshifted PAH emission lines (6.2,7.7 and 8.6um). We show statistically that the cluster galaxies have a lower S15um/S9um flux ratio than field galaxies, i.e., cluster galaxies already have lower star-formation activity at 0.9<z<1.7, pushing the formation epoch of these galaxy clusters to a higher redshift.