In this paper, we study the impact of aliased out-of-band interference signals on the information sum-rate of the maximum ratio combining receiver in massive multiple-input multiple-output (MIMO) uplink, with both perfect and imperfect channel estimates, in order to determine the required out-of-band attenuation in RF bandpass filters (BPFs). With imperfect channel estimates, our study reveals that as the number of base-station (BS) antennas ($M$) increases, the required attenuation at the BPFs increases as $\mathcal{O}(\sqrt{M})$ with $M \to \infty$, provided the desired information sum-rate (both in the presence and in the absence of AOOBIs (aliased out-of-band interferers)) remains fixed. This implies a practical limit on the number of BS antennas due to the increase in BPF design complexity and power consumption with increasing $M$.