The Baloney in Salon's Story About Memory Infections

The major online web site www.salon.com (Salon) has put up a story on human memory with the very goofy-sounding title, “A protein in your brain behaves like a virus, infecting your cells with memories.” The story is speculative nonsense, and it seems to be based on a BS hogwash misstatement (but not one told by the writer of the story). The story tells us, “Viruses may actually be responsible for the ability to form memories,” a claim that not one in 100 neuroscientists has ever made.

The story mentions some research done by Jason Shepherd, associate professor of neurobiology at the University of Utah Medical School. Shepherd has done some experiments testing whether a protein called Arc (also known as Arc/Arg 3.1) is a crucial requirement for memory formation in mice. He does this by removing the gene for Arc in some mice, which are called “Arc knockout” mice, meaning that they don't have the gene for Arc, and therefore don't have the Arc protein that requires the Arc gene.

In the Salon.com story, we read the following claim by Shepherd: “ 'If you take the Arc gene out of mice,' Shepherd explains, 'they don’t remember anything.' ” Right next to this quote, there's a link to a 2006 scientific paper by 26 scientists, none of them Shepherd (although strangely the Salon story suggests that the paper is research by Shepherd and his colleagues). Unfortunately, the paper does not at all match the claim Shepherd has made, for the paper demonstrates very substantial memory and learning in exactly these “Arc knockout” mice that were deprived of the Arc gene and the Arc protein.

The Morris water maze test is a test of learning and memory in rodents. In the test a rodent is put at the center of a large circular tub of water, which may or may not have milk powder added to make the water opaque. Near one edge of the tub is a submerged platform that the rodent can jump on to escape the tub. A rodent can find out the location of the platform by exploring around. Once the rodent has used the platform to escape the tub, the rodent can be put back in the center of the tub, to see how well the rodent remembers the location of the little platform previously discovered. 

In the paper by the 26 scientists, both 25 normal mice and 25 “Arc knockout” mice were tested on the Morris water maze. The “Arc knockout” mice remembered almost as well as the normal mice. The graphs below show the performance difference between the normal mice and the mutant “Arc knockout mice,” which was only minor. The authors state that in another version of the Morris water maze test of learning and memory, there was no difference between the performance of the “Arc knockout” mice and normal mice. We are told, “No differences were observed in the cue version of the task.”

Unimpressive “water maze” results from the paper by 26 scientists, showing no big effect from knocking out the Arc gene 

What title did the 26 scientists give their paper? They gave it the title, “Arc/Arg3.1 Is Essential for the Consolidation of Synaptic Plasticity and Memories.” But to the contrary, the Morris water maze part of the experiments shows exactly the opposite: that you can get rid of Arc/Arg3.1 in mutant “knockout” mice, and that they will still remember well, with good consolidation of memories (the test involved memory over several days, which requires consolidation of learned memories). The 26 authors have given us a very misleading title to their paper, suggesting it showed something, when it showed exactly the opposite.

A 2018 paper by 10 scientists testing exactly the same thing (whether “Arc knockout” mice do worse on the Morris water maze) found the same results as the paper by the 26 authors: that there were only minor differences in the performance of the mutant “Arc knockout mice” when they were tested with the Morris water maze test. In fact that paper specifically states, “Deletion of Arc/Arg3.1 in Adult Mice Impairs Spatial Memory but Not Learning,” which very much contradicts Shepherd's claim that “If you take the Arc gene out of mice, they don’t remember anything.”

Unimpressive “water maze” results from the 2018 paper by 10 scientists, showing no big effect from knocking out the Arc gene

So why did scientist Shepherd make the false claim that “If you take the Arc gene out of mice, they don’t remember anything,” a claim disproved by the paper by the 26 scientists (despite its misleading title contradicting its own findings), and also disproved by the 2018 paper? Shepherd has some explaining to do.

In the paper here, Shepherd and his co-authors state, “The neuronal gene Arc is essential for long-lasting information storage in the mammalian brain,” a claim that is inconsistent with the experimental results just described in two papers, which show long-lasting information retention in mice that had the Arc gene removed. In the paper Shepherd refers in a matter-of-fact manner to “information storage in the mammalian brain,” but indicates he doesn't have any real handle on how such a thing could work, by confessing that “we still lack a detailed molecular and cellular understanding of the processes involved,” which is the kind of thing people say when they don't have a good scientific basis for believing something.

We have seen here two cases of the misuse of the word "essential" in scientific papers.  I suspect that the word is being misused very frequently in biological literature, in cases where scientists say one thing is essential for something else when the second thing can exist quite substantially without the first.  Similarly, a scientific paper found that the phrase "necessary and sufficient" is being massively misused in biology papers that failed to find a "necessary and sufficient" relationship.  To say that one thing is necessary and sufficient for some other thing means that the second thing never occurs without the first, and the first thing (by itself) always produces the second.  An example of the correct use of the phrase is "permanent cessation of blood flow is both necessary and sufficient for death." But it seems biologists have massively misused this phrase, using it for cases in which no such "necessary and sufficient" relationship exists.  A preposterous example is the paper here which tells us in its title that two gut microbes (microorganisms too small to see) are "necessary and sufficient" for cognition in flies -- which is literally the claim that tiny microbes are all that a fly needs to think or understand. 

Postscript: I have received a reply from Jason Shepherd, who quite rightfully scolds me for misspelling his name in the original version of this post. I apologize for that careless error.  Here is Shepherd's reply, which I am grateful to receive:

"You've conflated two different terms and don't seem to understand what consolidation is: Learning and Memory. Arc KO mice can indeed learn information. Memory is the storage and retention of information that was learned. Arc KO mice have normal short term memory, as in if you test them in the water maze and other memory tasks they do, indeed, show memory. However, if you come back a day later...there is NO retention of that information. There is, in fact, an essential role for this gene in the consolidation of memory. That is, conversion of short to long-term memory."

I do not find this to be a satisfactory explanation at all, and when I wrote the post I was well aware of what consolidation is when referring to memory:  a kind of firming up of a newly acquired memory so that it is retained for a period of time such as days.  As for Shepherd's claim about Arc KO mice (mice that have had the Arc gene knocked out) that "if you come back a day later...there is NO retention of that information," that is clearly disproved by the two scientific papers that I cited.  The water maze test is a test of learning, conducted over 5 or more days,  showing a mouse's improved ability over multiple days to find a hidden platform. The graphs cited above very clearly show the retention and consolidation of learned information over multiple days in Arc knockout mice. The mice have got better and better at finding the hidden platform in the water maze, improving their score over several days, which they could only have done if they had retained information learned in previous days.  In three of the graphs above, we see steady improvement of Arc knockout mice in the water maze test, over a 9-day period. If it were actually true that in Arc KO mice "there is NO retention of that information," we would see a straight horizontal line in the performance graph for the water maze, not a diagonal line showing steady improvement from day-to-day.   

In his reply Shepherd rather seems to insinuate that the water-maze test is merely a test of short-term memory, but it is very well known that the Morris water-maze test is a test of long-term memory and memory consolidation.  For example, here we read that the Morris water maze test "requires the mice to learn to swim from any starting position to the escape platform, thereby acquiring a long-term memory of the platform’s spatial location."  And the link here tells us that the Morris water-maze tests "several stages of learning and memory (e.g., encoding, consolidation and retrieval)."  And the paper here refers to "spatial long-term memory retention in a water maze test."  And the recent Nature paper here says, "We next examined the hippocampus-dependent long-term memory formation in cKO mice by Morris water maze test." Figure 1 of the paper specifically cites a Morris water maze test, speaking just as if it tests what the paper calls "long-term memory consolidation." The first line of the scientific paper here states that the Morris water maze test was established to test things such as "long-term spatial memory." 

Here are some relevant comments from a science book about memory, comments disputing the idea that the Arc protein has anything to do with long-term memory persisting for days: 

"The half-life of the Arc protein is a few hours: think about the implications of this fact...Thus, if Arc (or any other protein with similar regulation and kinetics) participates in synaptic potentiation, this mechanism can only allow potentiation to be maintained for less than a day. Induction of Arc or any similar protein is clearly not a maintenance mechanism for very long-lasting events."

Post-postscript: I should have also included the graphs below from the 2018 paper on Arc knockout mice (from Figure 4).  In the first graph we see the performance of Arc knockout mice in the Morris water maze test, compared to normal mice. There is no noticeable difference in the 9-day test: the two lines match almost exactly -- just as if knocking out the Arc gene had produced no effect on memory or learning. 

The second and third graphs show "search strategies" used by Arc knockout mice (cKO) and the normal mice (WT-control).  The two graphs look the same. In his latest reply (which you can read below), Shepherd quotes this paper talking about a difference in search strategies between the two groups, but in the graphs above we fail to see any real  difference. The caption for these graphs says, "Similar progression of navigational strategy was observed in both groups."  Apparently there were two different experimental results, one of which failed to show any real difference in the search strategies used by the mice lacking the Arc gene. 

Post-post-postscript: In a 2020 paper Shepherd and his co-authors find that the Arc gene is "not required for LTP maintenance," a finding that seems to conflict with the claim that the Arc gene is vital for memory, or is at least discouraging to those who have made such a claim. The authors say they were "surprised by this result."