We present a 2.5σ detection of the Integrated Sachs-Wolfe (ISW) effect and discuss the constraints it places on cosmological parameters. We cross-correlate microwave temperature maps from the WMAP satellite with a 4000 deg 2 luminous red galaxy (LRG) overdensity map measured by the Sloan Digital Sky Survey. These galaxies have accurate photometric redshifts (∆z ∼ 0.03) and an approximately volume limited redshift distribution from z ∼ 0.2 to z ∼ 0.6 well suited to detecting the ISW effect. Accurate photometric redshifts allow us to perform a reliable auto-correlation analysis of the LRGs, eliminating the uncertainty in the galaxy bias, and combined with cross correlation signal, constrains cosmological parameters -in particular, the matter density. We use a minimum variance power spectrum estimator that optimally weights the data according to expected theoretical templates. We find a 2.5σ signal in the Ka, Q, V, and W WMAP bands, after combining the information from multipoles 2 ≤ l < 400. This is consistent with the expected amplitude of the ISW effect, but requires a lower matter density than is usually assumed: the amplitude, parametrized by the galaxy bias assuming ΩM = 0.3, ΩΛ = 0.7 and σ8 = 0.9, is bg = 4.05 ± 1.54 for V band, with similar results for the other bands. This should be compared to bg = 1.82 ± 0.02 from the auto-correlation analysis. These data provide only a weak confirmation (2.5σ) of dark energy, but provide a significant upper limit: ΩΛ = 0.80 +0.03 −0.06 (1σ) +0.05 −0.19 (2σ), assuming a cosmology with ΩM + ΩΛ = 1, Ω b = 0.05, and σ8 = 0.9, and w = −1. The weak cross-correlation signal rules out low matter density/high dark energy density universes and, in combination with other data, strongly constrains models with w < −1.3. We provide a simple prescription to incorporate these constraints into cosmological parameter estimation methods for (ΩM , σ8, w). We find no evidence for a systematic contamination of ISW signal, either from Galactic or extragalactic sources, but we do detect some large statistical fluctuations on smaller scales that could affect analyses without the template weighting.