Commentary Summary
This work extends the authors’ previous studies on the role of the peripheral immune system in AD. They have reported on the infiltration of peripheral monocytes through the blood-brain barrier (BBB), a significant mechanism of clearance of amyloid from the brain, and that peripheral macrophages from AD patients are defective in Aβ phagocytosis. These are important findings, and if other groups can confirm defects of Aβ uptake in Alzheimer patients’ macrophages, it may become a useful diagnostic assay. The present work further characterizes this defective phagocytosis to include defective intracellular trafficking of Aβ and phenotypic response to amyloid in AD brain tissue, as well as identifying a novel gene (MGAT) suggested to play a role in the defective phenotype.

The title and press releases surrounding this work emphasize the ability of a curcumin family compound (bisdesmethoxycurcumin, BDM-Curc) to correct the putative phagocytosis defect in AD. The structural characteristics of curcumin (Curc), enabling it to inhibit multiple aspects of...  Read more

Commentary Summary
This work extends the authors’ previous studies on the role of the peripheral immune system in AD. They have reported on the infiltration of peripheral monocytes through the blood-brain barrier (BBB), a significant mechanism of clearance of amyloid from the brain, and that peripheral macrophages from AD patients are defective in Aβ phagocytosis. These are important findings, and if other groups can confirm defects of Aβ uptake in Alzheimer patients’ macrophages, it may become a useful diagnostic assay. The present work further characterizes this defective phagocytosis to include defective intracellular trafficking of Aβ and phenotypic response to amyloid in AD brain tissue, as well as identifying a novel gene (MGAT) suggested to play a role in the defective phenotype.

The title and press releases surrounding this work emphasize the ability of a curcumin family compound (bisdesmethoxycurcumin, BDM-Curc) to correct the putative phagocytosis defect in AD. The structural characteristics of curcumin (Curc), enabling it to inhibit multiple aspects of AD pathogenesis, are not fully elucidated. BDM-Curc lacks the two phenolic methoxy groups found in curcumin, so knowing what these do could be important for developing new Curc-like drugs. Understanding mechanisms of Aβ uptake in human cells is likely important and the suggestion that MGAT may play a role would be a novel finding, if validated. However, factors other than an influence on phagocytic Aβ uptake (e.g., anti-oligomer aggregation, anti-inflammatory, antioxidant, etc.) appear to be crucial for curcumin’s full impact in AD models.

While the results in the title are provocative, it is unclear that the data are strong enough to support the conclusions. As discussed below, the data on differential effects of BDM-Curc (versus Curc) on phagocytosis, and the BDM-Curc impact on MGAT and TLR appear superficial and statistically underpowered. Further, there are numerous methodological issues which may affect interpretation of data, including the potent natural fluorescence of BDM-Curc, which could influence the endpoints measured and the accuracy of an Aβ-FITC phagocytosis assay to assess phagocytosis effects of compounds that may have Aβ-binding properties.

Introduction
Curcumin is known to have many anti-Alzheimer's properties (beyond those referenced in the current manuscript) including recruiting phagocytes to plaques in vivo and reducing Aβ-oligomer dependent lipid peroxidation (Frautschy et al., 2001), stimulating phagocytosis in vitro and in vivo (Cole et al., 2004), reducing protein oxidation and plaque burden in the Tg2576 model (Lim et al., 2001), exerting anti-Aβ aggregation in APP Tg2576 model (Yang et al., 2005) and in vitro (Ono et al., 2004), and the recent report of in vivo curcumin treatment for 7 days in a APPswe/PS1dE9 transgenic mouse leading to Curc labeling plaques, clearing plaques, reducing dystrophic neurites, and increasing soluble Aβ (Garcia-Alloza et al., 2007). It is therefore relevant to understand the structural aspects of the curcumin (Curc) molecule that mediate these different anti-Alzheimer’s effects, particularly in light of the ongoing clinical trial of Curc for AD at the UCLA AD Research Center.

BDM-Curc, which lacks methoxy groups on both phenolic rings, is interesting because its impact compared with that of Curc will help determine the importance of methoxy groups for these different anti-Alzheimer’s effects in structure-function studies. It is known that the lack of methoxy groups eliminates antioxidant activity conferred by the methoxy group at the ortho position. Since the phenolic groups in curcumin showed hydrogen bond acceptor properties, while those in BDM-Curc acted as hydrogen bond donors—explaining the differential polarity of these curcuminoids when mixed with various alcohols—it would have been more informative if the authors had shown data comparing the two curcuminoids. Both BDM-Curc and Curc have a diketone bridge that chelates metals, which may mediate some effects, but which also makes both compounds very unstable. Both are heavily glucuronidated in the intestine. Since BDM-Curc is a very minor component (2-3 percent) of semi-purified preparations in dietary supplements of Curc, it is unclear that it would play a major role in the plethora of published data on curcumin effects in animal models.

The conclusions in this manuscript are not well supported by the presented data for the following reasons:

The title might be somewhat misleading in that the effects were not observed in patients, but were entirely ex vivo. It is stated that AD and control patients in the UCLA ADRC (Alzheimer’s Disease Research Center) clinical trials are used to collect samples. Since these trials are ongoing and presumably blinded, one question that might arise or be discussed is whether any of the observed effects were due to the unknown test drugs that the patients were taking at the time of blood sampling. The title also implies that BDM-Curc effects are a primary conclusion of the paper, but only four of the 12 figures/tables evaluate BDM-Curc.

The authors state that active fractions from curcuminoids were isolated to identify the most immunostimulatory component, and it was found by LC/MS that this was BDM-Curc. Since this seems to be a major conclusion in title and abstract, it is somewhat puzzling that the data is not shown. It would be important to see the chromatograms, with internal standards to establish validity of the technique. More specifically, the authors state that for HPLC the retention time of BDM-Curc is 2.17 min measured by UV absorption at 220 nm. This is much shorter than what is reported in published data and seen in our independent observations where BDM-Curc elutes after Curc. The Methods state that “to verify the pharmacological activity, the minor curcumin, BDM-Curc, was also chemically synthesized and also showed great ability to enhance Aβ phagocytosis by human macrophages” (SI Fig. 7). However, this figure shows Aβ uptake in relation to MGAT mRNA. This may be a typographical error as the discussion in the text refers to SI Fig. 8, but even that figure does not show that BDM-Curc enhances Aβ phagocytosis more than other curcuminoids. The reported result that chromatographically purified BDM-Curc and chemically synthesized BDM-Curc have similarly optimal stimulation of phagocytosis was also not shown, It would also have been valuable to show a dose response with BDM-Curc, and to validate the "IOD method" of quantifying phagocytosis with an AB dose-response curve. In addition, it would have been more convincing if the numerous semi-quantitative statements about the phagocytic response of cultured cells (e.g., excellent, extremely efficient, minimal) were supported by methodological descriptions of these apparently subjective assessments.

The conclusions that BDM-Curc might enhance Aβ phagocytosis in human macrophages are based on FITC-Aβ-mediated uptake, as evaluated by confocal microscopy (supplemental material SI Fig. 8). Curcuminoids emit fluorescence in a wide spectrum from 475-650 nm, making it difficult to discriminate between FITC-labeled Aβ and the fluorescence from BDM-Curc which partitions into lipid and which may bind Aβ aggregates like other curcuminoids. In our experience, using a Cytofluor fluorescence reader (at wavelengths 485-590 nm), the fluorescence emitted by BDM-Curc is 25-fold more intense than Curc at high doses and fivefold at low doses. This suggests possible cooperative effects from BDM-Curc partitioning into Aβ aggregates or lipid-rich cellular microcompartments. Therefore, in order to make the conclusion that BDM-Curc increases Aβ aggregate phagocytosis better than Curc, controls with BDM-Curc alone and without unlabeled Aβ aggregates would be critical, as would dose response effects and comparisons of Curc with BDM-Curc. The dose needed is inconsistently reported: the text of two supplemental figure legends (SI Figs. 8 and 10) report that 0.1 mM was used, but the manuscript states 0.1 μM was used.

Without more evidence of anti-Alzheimer's properties of BDM-Curc, any potential relevance of BDM-Curc to Curc extract effects in animal models or epidemiology in India should be viewed cautiously for the following reasons: 1) Patients ingesting BDM-Curc in Curc extracts used in the published studies do not show free unglucuronidated levels; 2) Free BDM-Curc is highly unstable at pH 7.4; 3) Turmeric is ≤0.15 percent BDM-Curc, so even if BDM-Curc could be orally absorbed and was stable at pH 7.4, and not inactivated by glucuronidation, the authors’ proposed levels needed to affect phagocytosis could not be achieved in plasma by oral administration of the Curc extracts that have undergone previous testing for safety and efficacy.

MGAT (Table 1, Fig. 1, SI Fig. 7, SI Fig. 9)
The authors report microarray analysis (Table 1) showing 327-fold upregulation of MGAT3 mRNA in control macrophages treated with Aβ compared to AD macrophages treated with Aβ. It would be more informative to also see data on untreated cells, because the difference between AD macrophages and normal macrophages could be independent of the Aβ treatment. Only two cases per group (n = 2) were used in this microarray experiment, which is statistically dubious. Levels in expression of MGAT mRNA score in Aβ-stimulated macrophages from AD patients, and controls vary so widely within groups (six orders of magnitude, Fig. 1) that to prove this reflects biological activity would require additional data. If confirmed by more cases, a demonstration that MGAT played a role in phagocytosis would be a novel finding.

MGAT mRNA results, SI Fig. 7 (Supplemental Data) show a non-significant, weak positive correlation of MGAT mRNA and Aβ uptake. The likelihood of a correlation seems unconvincing and appears to depend on one data point (the one with low MGAT). It isn't clear which data points refer to macrophages taken from eight AD cases versus four control cases. It is important to show whether the observed changes in mRNA levels reflect changes in protein levels. Fig. 9 shows quantification of MGAT mRNA upregulation by BDM-Curc treatment (the graph is apparently mislabeled curcumin instead of BDM-Curc) in macrophages from AD patients; however, without showing the response of non-AD macrophages, the significance of these findings remains unclear.

It would be invaluable to investigate the relevance of the findings to AD brain: is there any evidence of changes in MGAT protein levels, enzyme activity levels, or mRNA expression?

The authors report that phagocytosis of Aβ in macrophages from one control subject is mediated by MGAT3 (based on silencing with siRNA). Since many pathways regulate phagocytosis, the data presented are not sufficient or thorough enough to make the relevance of this pathway convincing.

Data from AD slices (Fig. 6)
Surprisingly, for this article whose main point regards activity of BDM-Curc, the micrograph observations on AD brain slices with added macrophages did not include an assessment of the impact of BDM-Curc. In addition, some of the conclusions are poorly supported by data. For example, there is the suggestion that in AD brain slices, added AD macrophages uploaded then released Aβ, but no Aβ ELISAs of the media were done to check whether this occurred. Usually the AD brain slice/microglia (or macrophage) model would be used to show phagocytes accumulating around (non-phagocytosing) or inside (phagocytosing) Aβ-immunoreactive plaques (Frautschy et al., 2001; Cole et al., 2004). It is not clear why Aβ-ir plaques are not visualized with the anti-Aβ staining.

Toll-like Receptors (Fig. 4, SI Fig. 10)
Genetic knockout of TLR4 ([C3H/HeJ] TlrLPS-d) increases amyloid burden, thio S staining, as well as Aβ40 and Aβ42 in AppSw/PS1 Tg85 mice (Tahara et al., 2006). The BDM-Curc induction of TLR4 mRNA was observed in macrophages from one patient; this should be confirmed in other patients. Anti-TLR2-Phycoerythrin-detected TLR2 protein levels were reported on using flow cytometry of Aβ-treated cells (Supp Fig. 10) and fluorescence to detect changes in TLR2 (TLR3 and TLR4 protein levels are not reported). Anti-TLR2-Phycoerythrin was detected in the FL2-H channel of a flow cytometer that detects in the wavelength region 585 +/- 21 nm, which also overlaps with fluorescence spectrum of curcuminoids (see above). Therefore, in control macrophages, the increased fluorescence in BDM-Curc-treated cells could be due to BDM-Curc fluorescence and not due to anti-TLR-2-PE staining; controls of single stains are needed. The SI Fig. 10 (Supplemental Data) legend appears mislabeled since black appears to represent overlap, not BDM-Curc treatment. In panel B (no Aβ) the described increase with BDM-Curc is observable with data presented, but the non-specific labeling seems to be incorporated into the calculations (increase from 47 to 58 fluorescent intensity units).

It appears from flow cytometry that untreated cells are more TLR2-positive, compared to BDM-Curc-treated, in contrast to the stated result of an “increase” in median fluorescence with BDM-Curc (0.1 mM) from 45-65, which is hard to discern with the presented graph. Although it has previously been suggested and shown that some TLRs can enhance clearance, their upregulation is also likely to stimulate a cytokine profile that may aggravate AD pathogenesis, or directly exacerbate neurotoxicity. In fact, the TLR2/MyD88 pathway in microglia mediates neurodegeneration associated with bacterial stimulation of microglia (Lehnardt et al., 2006) and TLR4 may exacerbate neuron death associated with hypoxia-ischemia; TLR4-deficient mice are resistant to neurodegeneration caused by ischemia (Lehnardt et al., 2003). TLR2 is upregulated in ALS models (Nguyen, et al., 2001). Therefore, it is disappointing that the inflammatory cytokine profile was not presented. For Curc, it is unlikely that it works only by stimulating TLRs since it has such a pronounced impact on inhibiting inflammatory cytokines, the opposite effect of some TLR upregulation. In fact, Curc inhibits TLR4 signaling by preventing dimerization of TLR4 (Youn et al., 2006). Curc inhibits LPS-induced inflammation by binding the signaling-adaptor protein MD-2, which is a TLR4-binding protein and part of the endotoxin surface receptor complex (Gradisar et al., 2007).

Other
Fig. 6 legend has apparent typographical errors describing panels, stating that AD macrophages are C and D, but the panel shows it is E-H. Further inconsistencies in reported dyes are confusing: anti-Neun/Alexa594 is reported as red, but anti-Aβ/Alexa594 is reported as yellow, and anti-CD68/Alexa488 is reported as green in panels A, B, E, and F, but as yellow in panel C, as well as other indecipherable color reports in Fig. 6 legend. The “seven yellow cells” in 6E do not appear to be shrunken as the figure legend states.

Further, the description of Fig. 1 in the results section does not seem to agree with the figure itself. For example, the mean score of control subjects, +2.190, is a higher value than all but one of the samples, and the age-stratified mean for control subjects (+3.77) is higher than any value in the figure, and therefore cannot represent the actual mean.

View all comments by Sally A. Frautschy
View all comments by Bruce Teter

Reply to Frautschy, Teter Comment

In response to the comments by Frautschy and others, the objectives of our paper are first to explain the immune mechanisms of amyloidosis in Alzheimer disease patients and second to find out what can be done about clearance of amyloidosis from the patient’s brain. The emerging answers are that amyloidosis is contributed by insufficient clearance by the Alzheimer patients’ innate immune system and that modulation of the innate immune system has positive effects on amyloid-β clearance.

There is no problem in distinguishing FITC-amyloid-β by fluorescence microscopy from curcuminoids, which (at 0.1 microM) are not visible by fluorescence microscopy. Amyloid-β is also revealed by immunostaining with amyloid-β antibody or by electron microscopy. This can be seen in the pictures of FITC-Aβ in Figs. 2, 3, 5 in the current PNAS publication (1) or the Figs. 2 and 3 (using anti-Aβ immunofluorescence or electron microscopy) in our previous publication (2). The responses of individual patients and clinical data correlations were...  Read more

Reply to Frautschy, Teter Comment

In response to the comments by Frautschy and others, the objectives of our paper are first to explain the immune mechanisms of amyloidosis in Alzheimer disease patients and second to find out what can be done about clearance of amyloidosis from the patient’s brain. The emerging answers are that amyloidosis is contributed by insufficient clearance by the Alzheimer patients’ innate immune system and that modulation of the innate immune system has positive effects on amyloid-β clearance.

There is no problem in distinguishing FITC-amyloid-β by fluorescence microscopy from curcuminoids, which (at 0.1 microM) are not visible by fluorescence microscopy. Amyloid-β is also revealed by immunostaining with amyloid-β antibody or by electron microscopy. This can be seen in the pictures of FITC-Aβ in Figs. 2, 3, 5 in the current PNAS publication (1) or the Figs. 2 and 3 (using anti-Aβ immunofluorescence or electron microscopy) in our previous publication (2). The responses of individual patients and clinical data correlations were examined in a previous publication (3).

Our work relates to human tissues and blood cells from patients with Alzheimer disease, which makes a direct comparison with transgenic animals (which do not have a specific immune defect) difficult. We performed the studies in macrophages and monocytes from Alzheimer patients over a 6-year period. Logistically, it is difficult to ask that the blood specimens of over 140 patients collected over a 6-year period would be analyzed by the same techniques that were developed later during the course of this study. However, the immune defects in phagocytosis have been observed by fluorescence microscopy in a majority of patients and the biochemical defects in a small number of patients (MGAT3 in over 20 patients, TLR defects in four patients) but with a remarkable consistency. Without doubt, many factors might affect the immune system, including drugs, hormones, stress, infection, etc. However, the patients in Phase 1 – preclinical study (where we are now) have to be examined, as they present themselves. Phase 2 and 3 studies will be possible at a later stage of investigation.

Bisdemethoxycurcumin showed greatest effect on phagocytosis when compared to unfractionated curcuminoids or other fractions. In order to obtain the most reproducible results, we chose to work with a pure chemical, bisdemethoxycurcumin, not the unfractionated material.

Regarding MGAT3 and TLR results, the data, such as Fig. 4, speak for themselves since they are consistent (patients vs. controls). Recent results continue to support the conclusions about transcriptional effects of curcuminoids. The studies of MGAT3 protein levels in the brain are difficult since a good antibody is not available. We agree that more flow cytometric testing of TLR proteins in PBMCs treated with bisdemethoxycurcumin is warranted (the legend had an error and we plan to correct this).

The results of PBMC clearance of Aβ in brain sections in Fig. 6 are striking and deserve closer scrutiny of the legends. Similar results have been obtained in at least five other experiments. The tissues were obtained from the UCLA brain bank. We have not seen the effects on neuritic plaques in these frozen tissues, but further work is ongoing.

Our intention in this study was to identify and characterize the most potent anti-Alzheimer disease agent in mixture of curcuminoids, not to study curcumin SAR. In fact, bisdemethoxycurcumin does possess antioxidant activity (4). In the AAPH-induced linoleic acid antioxidation test or the DPPH-radical scavenging test, bisdemethoxycurcumin does possess significant activity as an antioxidant.

We are aware of the chemical properties of curcumins and related materials, and the compounds are readily handled with proper care. All of the compounds were fully characterized spectrally, and instability was not a problem during our analytical and synthetic studies. We are aware of the metabolic properties of curcumins and in fact “Dynamic Medicinal Chemistry” has been a major effort leading our studies in this and other areas (5). Our intention was not to study bisdemethoxycurcumin in a pharmaceutical sense. For our purposes, of greatest relevance was the apparent effective concentration of the active pharmacological agent at the target site and not the relative percent of material in dietary supplements.

The purpose of the study was to biotrack the most pharmacologically active constituent. That no curcumin was present in the final HPLC in the purification of bisdemethoxycurcumin was not surprising and speaks to the alacrity of our separation approach. Agreeably, the retention time was short, but the solvent polarity gradient was steep and quite effective. Of course, elution profiles are a function of the matrix employed and the history of the matrix. To confirm the activity of bisdemethoxycurcumin, indeed, synthetic bisdemethoxycurcumin was prepared and fully characterized spectrally. That synthetic bisdemethoxycurcumin is also highly active supports the exciting observation that the minor constituent in curcuminoids contains remarkable biological properties.

It was beyond the scope of this study to examine the glucuronidation of bisdemethoxycurcumin, but it is important to point out that 1) glucuronides undergo enterohepatic cycling, and urinary metabolite levels may not reflect metabolic disposition in the blood; 2) bisdemethoxycurcumin may not be “free” in the biological context but in fact associated with proteins, thus confounding apparent observations about solution stability study data; and 3) the effective concentration or accumulation of bisdemethoxycurcumin in the target tissue or cell may be much different than that estimated from plasma levels. The examination of these points is the subject of additional studies.

References:
1. Fiala M, Liu PT, Espinosa-Jeffrey A, Rosenthal MJ, Bernard G, Ringman JM, Sayre J, Zhang L, Zaghi J, Dejbakhsh S, Chiang B, Hui J, Mahanian M, Baghaee A, Hong P, Cashman J. Innate immunity and transcription of MGAT-III and Toll-like receptors in Alzheimer's disease patients are improved by bisdemethoxycurcumin. Proc Natl Acad Sci U S A. 2007 Jul 31;104(31):12849-54. Epub 2007 Jul 24. Abstract

2. Fiala M, Lin J, Ringman J, Kermani-Arab V, Tsao G, Patel A, Lossinsky AS, Graves MC, Gustavson A, Sayre J, Sofroni E, Suarez T, Chiappelli F, Bernard G. Ineffective phagocytosis of amyloid-beta by macrophages of Alzheimer's disease patients. J Alzheimers Dis. 2005 Jun;7(3):221-32; discussion 255-62. Abstract

3. Zhang L, Fiala M, Cashman J, Sayre J, Espinosa A, Mahanian M, Zaghi J, Badmaev V, Graves MC, Bernard G, Rosenthal M. Curcuminoids enhance amyloid-beta uptake by macrophages of Alzheimer's disease patients. J Alzheimers Dis. 2006 Sep;10(1):1-7. Abstract

4. Somparn P, Phisalaphong C, Nakornchai S, Unchern S, Morales NP. Comparative antioxidant activities of curcumin and its demethoxy and hydrogenated derivatives. Biol Pharm Bull. 2007 Jan;30(1):74-8. Abstract

5. Cashman JR, MacDougall JM. Dynamic medicinal chemistry in the elaboration of morphine-6-glucuronide analogs. Curr Top Med Chem. 2005;5(6):585-94. Review. Abstract

View all comments by Milan Fiala