We measure equivalent widths (EW) - focussing on two unique features (NaI and TiO2) of low-mass stars (\leq 0.3 M\odot) - for luminous red galaxy spectra from the the Sloan Digital Sky Survey (SDSS) and X-Shooter Lens Survey (XLENS) in order to study the low-mass end of the IMF. We compare these EWs to those derived from simple stellar population models computed with different IMFs, ages,[{\alpha}/Fe], and elemental abundances. We find that models are able to simultaneously reproduce the observed NaD {\lambda}5895 and NaI {\lambda}8190 features for the lower-mass (~{\sigma}*) ETGs but deviate increasingly for more massive ETGs, due do strongly mismatching NaD EWs. The TiO2 {\lambda}6230 feature and the NaI {\lambda}8190 feature, may be a powerful IMF diagnostic, with age and metallicity effects orthogonal to the effect of IMF on the feature's strength. We find that both features correlate strongly with galaxy velocity dispersion. The XLENS ETG (SDSSJ0912+0029) and one SDSS ETG (SDSSJ0041-0914) appear to require both an extreme dwarf-rich IMFs and a high sodium enhancement ([Na/Fe]=+0.4). However, lensing constraints on the total mass of the XLENS system within its Einstein radius limit a bottom-heavy IMF with a power-law slope to x \leq 3.0 at the 90% C.L. We conclude that NaI and TiO features, in comparison with state-of-the-art SSP models, suggest a mildly steepening IMF from Salpeter to x~3.0 for ETGs in the range {\sigma}=200-335 km/s.