We study the content in S-bearing molecules of protoplanetary disks around low-mass stars. We used the new IRAM 30-m receiver EMIR to perform simultaneous observations of the $1_{10}-1_{01}$ line of H$_2$S at 168.8 GHz and $2_{23}-1_{12}$ line of SO at 99.3 GHz. We compared the observational results with predictions coming from the astrochemical code NAUTILUS, which has been adapted to protoplanetary disks. The data were analyzed together with existing CS J=3-2 observations. We fail to detect the SO and H$_2$S lines, although CS is detected in LkCa15, DM\,Tau, and GO\,Tau but not in MWC\,480. However, our new upper limits are significantly better than previous ones and allow us to put some interesting constraints on the sulfur chemistry. Our best modeling of disks is obtained for a C/O ratio of 1.2, starting from initial cloud conditions of H density of $2\times 10^5$ cm$^{-3}$ and age of $10^6$ yr. The results agree with the CS data and are compatible with the SO upper limits, but fail to reproduce the H$_2$S upper limits. The predicted H$_2$S column densities are too high by at least one order of magnitude. H$_2$S may remain locked onto grain surfaces and react with other species, thereby preventing the desorption of H$_2$S.Read less <