Here we demonstrate that the dimer/trimer fraction of a naturally secreted form of human Aβ is dramatically more potent in inducing cofilin-actin rods than traditionally prepared human Aβsyn oligomers. We estimate Aβd/t activity is about 4000 fold greater than traditionally prepared human Aβsyn oligomers, but this is based upon an indirect quantification of dimer/trimer concentration using the dot blot to determine total Aβ and gel quantification to estimate the dimer/trimer as a percentage of the total. The maximal activity for Aβd/t (1.1 ng/mL or ~250 pM based on monomer equivalents) is within its estimated physiological concentrations based on the amount extracted from human AD brain [21, 22] and the amount detected in cerebrospinal fluid from normal and AD patients . Because the rod-inducing activity of the Aβd/t fraction is neutralized with an Aβ antibody, it is Aβ that appears to be the active rod-inducing component. Furthermore, the lack of significant rod-inducing activity of the Aβ monomer fraction from the 7PA2 medium, even at concentrations three-fold higher than Aβd/t, suggests the SDS-stable dimer is the minimal active unit in promoting cofilin-actin rod formation.
It is not known what is responsible for the size distribution of species within the monomer, and SDS-stable low-n oligomer pools. However, naturally occurring human Aβ species modified at their N-terminus to generate both 3-42 and 4-42 isoforms have been reported . The 3-42 isoform with an N-terminal glutamate is capable of forming the pyroglutamate form, which along with 1-40, 1-42 and 4-42 isoforms, are among the most abundant in control, familial AD and sporadic AD brain . Thus mixed oligomers of these different species probably account for the size distribution observed on Western blots.
The nature of the cross-links to generate the SDS-stable species has also been questioned. It was previously shown that tyrosine10 in the human Aβ sequence can undergo oxidation in the presence of Cu2+ and peroxide to form a specific di-tyrosine cross-link, which did not form in the absence of Cu2+ (or Fe3+) . However, other peroxide-induced dimers were produced but not characterized, and these presumably could also result in SDS-stable trimer and tetramer species. Rodent Aβ contains no tyrosine (F10 in place of Y10; Figure 3D), is deficient in formation of SDS-stable oligomers [46, 63], and does not induce cofilin-actin rods when used at 1 μM, even after Cu2+/peroxide treatment (Figure 3B), suggesting that tyrosine oxidized Aβ and the formation of stable dimers may account for enhanced activity of human Aβ. We investigated the biological activity of oxidized human Aβ prepared with peroxide in the presence and absence of copper [46, 47]. Both oxidized human Aβ preparations had greatly enhanced rod-inducing biological activity, with the one made in the presence of copper being active at a concentration within 10 fold of that of Aβd/t from 7PA2 cells. However, the enhanced activity of the human Aβ oxidized in the absence of Cu2+ suggests that species of Aβ dimer (or other SDS-stable low-n oligomers) other than those with a di-tyrosine cross-link also contribute to rod-inducing activity.
We assume that the lack of dimers in traditionally prepared oligomers of human Aβ arises from their rapid assembly at the high concentration of human Aβsyn used , combined with the antioxidant effects of DMSO, which is used to solubilize the peptide [44, 45]. Our data strongly supports the presence of oxidized human Aβ dimers as contributing to the potent rod-inducing activity of the 7PA2 Aβd/t fractions.
We found that the amount of the Aβd/t did not change in the medium over 3 d. This is due in part to the fact that rodent Aβ does not form SDS-stable oligomers and hence the continued secretion of rodent Aβ by the cultured neurons, which is stimulated by the presence of the human Aβd/t , does not contribute to the Aβd/t pool. It also suggests that neuronal uptake of the Aβd/t, which has been reported , occurs at a slow enough rate to not significantly deplete the extracellular pool during the 3 day exposure.
Enhancing cofilin's F-actin binding activity by overexpressing either of the cofilin phosphatases, slingshot or chronophin, enhanced rod formation in dissociated neurons in the absence of Aβ-treatment and expression of the cofilin kinase LIMK1 inhibited formation of rods even after Aβ-treatment. This result was not unexpected since overexpression of cofilin, especially the non-phosphorylatable cofilin S3A mutant, enhances rod formation . Expression of LIMK1 in organotypic slices reduced Aβ-mediated rod formation similar to its effect in dissociated neurons. The expression of cofilin phosphatases in organotypic slice cultures produced no significant increase in rod formation either alone or in response to Aβd/t. Although we find about a 70% adenoviral infectivity rate of cells in the slices, it is quite likely that the non-neuronal cells infect better than the neurons; thus for some adenoviruses the neuronal infectivity may be low enough that we would have difficulty in obtaining enough infected neurons to observe significant changes in rod formation.
Single infusions into the adult rat brain of either Aβ dimer extracted from human AD brain or Aβd/t at the identical concentration (250 pM) used in our studies caused transient memory and learning deficits when measured starting 2 h after infusion, and completed within a 2 h maximum time frame [20, 21]. Memory and learning deficits disappeared 24 h after the single infusions. It is worthy to note in this regard that the Aβd/t fraction induces rods in a statistically significant number of neurons by 2 h after treatment (about 25% of the maximal response) and that the Aβd/t rods are reversible, disappearing by 24 h after washout. Thus, formation and disappearance of rods in cultured neurons and organotypic slices correlate well with the changes observed in memory and learning in whole animals exposed to a single infusion of Aβd/t.
Nevertheless, the time of rod formation in response to Aβd/t treatment does not correlate well with acute Aβd/t effects on slice electrophysiology. Decreased long-term potentiation (LTP) and enhanced long-term depression (LTD) occur within 20 min of treating hippocampal slices in culture either with fractions containing Aβd/t  or with Aβ dimer extracted from human AD brain . This response is more rapid than the 2 h it takes to obtain a significant increase in rods in organotypic slices exposed to the Aβd/t fraction. Rod formation has been observed to occur within 10 min in organotypic slice cultures responding to anoxia or ATP-depletion [27, 29]. However, the rate of Aβd/t induced rod formation in organotypic slices maintained in neurobasal/B27 medium may be significantly slower than in acute slices prepared for electrophysiology and maintained in artificial cerebrospinal fluid (aCSF) owing to 0.6 μM insulin in the neurobasal/B27 medium not present in aCSF. Insulin helps neurons resist the pathogenic changes in cytoskeletal organization induced by Aβ . Alternatively or in addition, we suspect that cofilin dephosphorylation and altered actin dynamics in response to Aβd/t is rapid and precedes rod formation in some compartments, such as dendritic spines. Synaptic activity depends directly on cofilin function in regulating actin dynamics and may reflect early changes in plasticity [43, 68–70]. Thus, the LTP/LTD response to Aβd/t could be independent of rod formation yet the result of a localized change in cofilin activity. Significantly, wash out of Aβ after the electrophysiological changes have occurred did not result in any reversal in the altered LTP/LTD over 2 h. This persistent effect could arise within the organotypic slice from relatively tight binding of the Aβd/t to specific sites for which some evidence does exist [23, 67]. Alternatively, rod formation may have occurred by the time washout was initiated and rods could be responsible for the lack of rapid recovery. Rods sequester most of the cofilin [27, 33] needed to re-establish the balance in spine actin dynamics  and their formation would resist rapid recoveries of the spine cofilin pool.
The most sensitive pool of neurons forming rods in response to both synthetic Aβ oligomers and Aβd/t are those within the polymorphic/hilar region of the dentate gyrus along the mossy fiber tract into the CA3. The dentate gyrus is considered to play a central role in associative memory . Its major input comes via the perforant pathway with axons representing approximately one million excitatory entorhinal neurons from layer II. These axons end preferentially within the outer two thirds of the superficial molecular layer, mainly on the apical dendrites of the granule cells, but also on dendrites of interneurons . Cholinergic neurons from the basal forebrain provide another important afferent input, and also synapse with neurons of the dentate gyrus inner molecular layer. The CA3 pyramidal cells receive the granule cell output via the mossy fibers (granule cell axons) and aid in pattern completion . Because of its central role in associative memory, the dentate gyrus has been extensively studied in AD brain . There is an early loss of synapses (48% decrease in synapse to neuron ratio) before significant loss of neurons [42, 73, 74]. Based upon the degree of immunofluorescence labeling for the synaptic marker synaptophysin, there is a direct correlation of synaptic loss during AD progression. Early, mild and severe AD cases are accompanied by a decline in synaptophysin staining of about 25, 45 and 65%, respectively, in the outer and middle third of the molecular layer, with little to no loss in the inner third . Injection of single doses of Aβd/t-containing medium into brains of adult rats also leads to defects in associative memory and to memory consolidation with a striking inhibition of the synaptic increases that occur during memory consolidation in the dentate gyrus .