To determine the differences in NF-κB activation at both the single-cell and population level, we began by measuring time courses of p65-dsRed nuclear localization in response to each of three preparations – Sigma #L4524 from E. coli
055:B5 (Sigma), Invitrogen LPS-EB from E. coli 0111:B4 (EB) and a more highly purified form of EB called Ultrapure (UP) – at a single concentration (0.5 µg/mL). The concentration was chosen because it is common to much of the published literature, as well as our own work 
First, we seeded cells onto glass-coverslip 96-well plates. These cells express a p65-dsRed fusion protein, driven by the endogenous mouse p65 promoter, as well as H2B-GFP under control of the ubiquitin promoter (to facilitate cell segmentation and tracking). We then simultaneously stimulated each well with one of the described LPS preparations, on an incubated epifluorescent microscope. Upon stimulation, active p65 fusion proteins shuttle into the nucleus, leading to changes in nuclear fluorescent intensity. LPS response dynamics were measured by the change in nuclear fluorescent intensity over time, normalized by the cytoplasmic intensity prior to stimulation.
We found that the different preparations led to significantly different activation dynamics (). For example, cells stimulated with Sigma exhibited a 40–55% longer average localization than cells stimulated with EB or UP. We also noted that Sigma-stimulated cells had a marked qualitative heterogeneity in the response at the single-cell level, with both transiently- and persistently-activated cells, as we have previously described 
. Cells stimulated with EB exhibited 51% lower-intensity activation and a three-fold longer time-to-peak on average, and were more likely to have subsequent strong secondary nuclear activation. The UP-stimulated cells had the highest peak intensity, about 17% higher than for Sigma, as well as a 45% longer time-to-peak.
Comparison of single-cell NF-κB activation dynamics for three different LPS preparations.
A possible explanation of these observations is that the difference in activation is simply due to a potency effect – for example, a given concentration of EB is equivalent to a lower concentration of Sigma and UP. To test this hypothesis and further compare the three LPS preparations, we determined the range of LPS concentrations that activate cells for each preparation. We treated our cells with varying concentrations of Sigma, EB and UP, spanning nine orders of magnitude in concentration, and determined the nuclear localization over time for roughly 100 cells in at least three experiments at each concentration and preparation.
We first attempted to determine the potency range in terms of the fraction of cells that were visibly activated. shows the active cell fraction over the entire concentration range for each LPS preparation. We found that the resulting activation profiles were strikingly different for each preparation. Specifically, Sigma was maximally active beginning at concentrations of 5·10−4 to 5·10−3 µg/mL, while UP reached its peak fractional activation at 0.5–5 µg/mL, 2–3 orders of magnitude higher. UP LPS was also the only preparation that was essentially ineffective at the lowest concentrations we tested. Surprisingly, even though EB activation exhibited the lowest intensity in , almost all of the EB-stimulated cells showed detectable NF-κB nuclear translocation, even at the lowest concentrations we used.
The potency window for each of the LPS preparations.
Comparison of the dynamics of the LPS response for each preparation is complicated by this broad difference in the activity range. As a result, we compared the activation dynamics for each preparation at all concentrations. The average and selected single-cell time courses for cells stimulated with LPS at various preparations and concentrations are shown in . As before, we found that the time-to-peak was generally longest for EB-stimulated cells, and depended on concentration for both Sigma- and UP-stimulated cells (). Furthermore, the peak intensity was comparable for all treatments, with the exception of UP-stimulated cells, for which intensity increased with concentration (). We also found an inverse relationship between peak amplitude and time-to-peak for UP-stimulated cells, similar to cells treated with TNF 
. In contrast, cells stimulated with Sigma generally maintained the same peak amplitude regardless of changes in time-to-peak ().
Activation dynamics for each of the LPS preparations and several concentrations and summary statistical comparisons.
We then compared aggregate activation dynamics across all preparations and concentrations by determining the cosine distance between each pair of average time courses (). The average time courses that are most similar have the shortest distance between their corresponding vectors. As expected, similarity between time courses often corresponds to cells treated with the same preparation of LPS and at similar concentrations. One interesting exception involves the UP preparation, for which the 0.05 µg/mL concentration produced an activation time course that was similar to many of the EB time courses. The similarity results from the sharp increase in average time-to-peak from the UP 5 µg/mL time courses to the UP 0.05 µg/mL time courses. The increase, from 52 minutes to 69 minutes to 129 minutes as concentration is decreased along this range, leads to a time-to-peak that is closer to the average EB time-to-peak across concentrations. With that exception, these findings do not support the hypothesis that the differences in NF-κB activation between preparations are related to a potency effect.
The similarity between average activation profiles across all concentrations and preparations.
Finally, we wondered how TNF secretion shaped the dynamics of NF-κB for each preparation of LPS. In our earlier work we found that cells treated with 0.5 µg/mL Sigma LPS had significantly different aggregate behavior when soluble TNF receptor II (sTNFRII) was added to the medium 
. In particular, the number of persistent cells was greatly reduced, suggesting that persistent activation depended on secreted TNF. We wondered if such sTNFRII sensitivity was a common feature of the LPS preparations. Accordingly, we treated cells simultaneously with sTNFRII and Sigma, EB or UP LPS over the same concentration range, and compared average time courses of NF-κB nuclear localization in the presence or absence of sTNFRII (). We also calculated the cosine distance between the curves of average NF-κB intensity for active cells with or without sTNFRII (). The responses of cells treated with high concentrations of Sigma were much more likely to be affected by sTNFRII, especially at early times (<100 minutes), than the responses of cells exposed to EB or UP (). These results suggest that an important phenotype and commonly measured phenotype, namely secretion of a cytokine, can vary depending on the nature of the LPS preparation. We also performed high-sensitivity ELISA at multiple timepoints from 0 to 3 hours, but interestingly, detected no TNF in the media for any of the LPS preparations (<10 pg/mL, data not shown).
Blocking paracrine signaling by TNF across all concentrations and preparations.
The effect of blocking TNF on the NF-κB activation time courses as a function of concentration and preparation.