Monday, January 30, 2023

Can’t just “get over it”: Long COVID

Many people have experienced lingering problems after recovery from COVID infection, a condition known as long COVID or post-COVID.  A Norwegian study found that symptoms persisted for 6 months in most patients (189 of 312), including most young adults.  (‘Long COVID’ preferred because ‘post-COVID’ is ambiguous.) 

What is long COVID? There are many symptoms, led by fatigue, loss of smell/taste, breathlessness, and cognitive impairment, with no obvious common cause or relationship. Increased blood clotting (prothrombotic) has been suspected, suggested by the involvement of the receptor for the COVID-19 spike protein, ACE2, but strong evidence and mechanisms have been lacking.  

These authors tested the blood of 21 patients with post-COVID syndrome (PCS), averaging nearly 2 years after onset of infection. They modeled real blood flow through vessels by collecting blood samples (treated with the anti-coagulant citrate) and sending it through narrow tubes coated with particular proteins. They looked at the binding of platelets, which are abundant cellular products that initiate blood clots upon being triggered by binding collagen. They found striking increases in platelets binding to collagen (shown, Figure 1a top panel). Antibodies against von Willebrand factor (VWF) produced equivalent binding though apparently with different patterns (middle panels). 

Figure 1A: Binding of platelets (yellow) from healthy control blood (left) or PCS blood (right) to collagen (top), anti-VWF (middle), or VWF (bottom).

Although these intriguing findings await confirmation (by others) and follow-up, of course, they are of the utmost importance given the enormous impact of COVID.  

Constantinescu-Bercu A, Kessler A, de Groot R, Dragunaite B, Heightman M, Hillman T, Price LC, Brennan E, Sivera R, Vanhoorelbeke K, Singh D, Scully M. Analysis of thrombogenicity under flow reveals new insights into the prothrombotic state of patients with post-COVID syndrome. J Thromb Haemost. 2023 Jan;21(1):94-100. doi: 10.1016/j.jtha.2022.10.013. Epub 2022 Dec 22. PMID: 36695401; PMCID: PMC9773628.  

Monday, December 26, 2022

Identifying Mitochondrial Functions by ‘Multiomic’ Profiling

Mitochondria make ATP through oxidative phosphorylation, thus providing energy for nearly all cellular functions. Many human disorders are attributed to mitochondrial dysfunction. Their functions seem narrow and their genomes, known for decades, were reduced to encoding a mere 13 proteins after transferring most to the nuclear genome. However, these authors note that ‘hundreds of mitochondrial proteins lack clear functions’. They previously (Stefely 2016) applied mass spectroscopy (MS) ‘multiomics’ to assign functions to mitochondrial uncharacterized (x) proteins (MXPs) in yeast. Here, they generated using CRISPR over 200 knockout (KO) cell lines, targeting 50 nuclear genes encoded MXPs plus 66 with known functions, and assessed in each line over 8,000 proteins, over 3,000 lipids and over 200 metabolites by MS (epic undertaking!). They found high reproducibility and dynamic range, with ‘many molecules showing regulation over 3-4 orders of magnitude’.

Fig 2e. Relative protein abundance in SLC30A9 KO cells compared to WT (“wild type”, i.e., normal) cells versus statistical significance with noted mitochondrial ribosome (black), OxPhos (blue), and mtDNA-encoded (red) proteins.

Some assessments confirmed expectations or were mild surprises, e.g., the importance of ALDH18A1 or NADK1 in proline synthesis (Fig 2a). Others revealed ‘new biology’, such as a key role for the putative zinc transporter SLC30A9 in mitochondrial ribosome and OxPhos proteins (Fig 2e, shown). They also found that one ’upstream (open) reading frame’ (PYURF) is a chaperone essential for complex I and coQ synthesis, linked a transporter (SLC30A9) to ribosomes, and found a second gene (RAB5IF) contributing to developmental disorders. They offer their “8.3 million distinct biomolecule measurements” online to help others ascribe additional functions, a promising resource.

Rensvold JW, Shishkova E, Sverchkov Y, Miller IJ, Cetinkaya A, Pyle A, Manicki M, Brademan DR, Alanay Y, Raiman J, Jochem A, Hutchins PD, Peters SR, Linke V, Overmyer KA, Salome AZ, Hebert AS, Vincent CE, Kwiecien NW, Rush MJP, Westphall MS, Craven M, Akarsu NA, Taylor RW, Coon JJ, Pagliarini DJ. Defining mitochondrial protein functions through deep multiomic profiling. Nature. 2022 Jun;606(7913):382-388. doi: 10.1038/s41586-022-04765-3. Epub 2022 May 25. PMID: 35614220; PMCID: PMC9310563.
NB PubPeer comment raises concerns regarding the methods, interpretations, and conclusions. 

Friday, December 16, 2022

Seasonal flu vaccination: short term protection, long term risk?

Influenza (flu) sickens millions and kills many thousands of Americans each year. Vaccination is intended to reduce the number cases and severity of illness. The flu virus changes its coat proteins each year, a ‘shape-shifting’ behavior that challenges the timely production of vaccines that are effective against flu variants. The US CDC evaluates vaccine effectiveness (VE) in thousands of outpatient respiratory illness patients, usually finding substantial vaccine protection with little risk Consequently, its Advisory Committee on Immunization Practices (ACIP) as well as the World Health Organization (WHO) recommend annual immunization of everyone over 6 months old unless contraindicated.
Here, the authors surveyed vaccination over 10 years among several thousand Japanese school children and several hundred school staff adults. Of the many forms of flu vaccine, Japan uses a quadrivalent (4 strains; trivalent prior to 2014) “split” vaccine based on influenza hemagglutinin (HA). “Split” vaccine means the virus was ‘disrupted’ by detergent (the equivalent of ‘heat killed’; viruses aren’t alive). The flu viruses used to prepare the vaccine are grown in eggs; alternative quadrivalent preparations are available for those with allergies to eggs, including Flublok and Flucelvax.

They found that morbidity was reduced in vaccinated elementary school students but elevated in middle school students (Fig 1). Most people who had been vaccinated ‘from infancy’ were also vaccinated in the 2019-2020 season (Fig. 2). Moreover, they “found that morbidity was significantly higher among elementary (P < 0.001) and middle (P < 0.05) school students who had been vaccinated since infancy than among those who had not been vaccinated since infancy” (Figure 3, shown).

Fig. 3 Relationship between morbidity and vaccination from infancy.

These data are self-reported via questionnaires, and therefore extra subjective. The authors propose no mechanism for how annual vaccination could cause increased morbidity. It seems probable that the association is not direct, not causative, but indirect through other behavioral or health status factors. This is an intriguing finding that should be analyzed and study that should be repeated.

Kajiume T, Mukai S, Toyota N, Kanazawa I, Kato A, Akimoto E, Shirakawa T. Effectiveness of seasonal influenza vaccine in elementary and middle schools: a 10-year follow-up investigation. BMC Infect Dis. 2022 Dec 6;22(1):909. doi: 10.1186/s12879-022-07898-y. PMID: 36474168; PMCID: PMC9724312.

Wednesday, November 16, 2022

Protein folding by AI: wrinkles

Tech giants Alpha and Meta (Google and Facebook) applied their Artificial Intelligence (AI) to fold proteins computationally, predicting 3-dimensional shapes from the 1-dimensional sequence data. Meta’s paper is still paywalled (preprint) but AlphaFold’s Nature papers from last year are available (Jumper, Tunyasuvunakool). 

The AlphaFold authors noted that the ~100,000 protein structures determined by conventional experimental means are a small portion of the “billions” extant in nature. Previous approaches “focus on either the physical interactions or the evolutionary history”, which they say relies on the availability of close homologues or works for (only) a few, small proteins and is otherwise “computationally intractable” (too hard). They evaluated in the 87 protein domains comprising the 14th Critical Assessment of (protein) Structure Prediction (CASP14) dataset, structures not yet deposited in the public Protein Data Bank (PDB). This permits a ‘blind’ (apriori) comparison of AI methods, by comparing their predictions with the newly-solved structures. 

Fig 1. a. Scores. b. Backbone. c. Side chains
 By this measure, AlphaFold is much better than its competitors (Fig 1a, shown, predicted vs experimental). It gains accuracy on backbone and side chains (1b, c) “by incorporating novel neural network architectures and training procedures based on the evolutionary, physical and geometric constraints of protein structures”. Using “multiple sequence alignments (MSAs) and pairwise” comparisons, it “predicts the 3D coordinates of all heavy atoms for a given protein using the primary amino acid sequence and aligned sequences of homologues as inputs”. So give it a bunch of similar sequences and structures and voila! it gives the ‘new’ one. Thy describe the process and you can download to code to inspect, modify, run yourself (open source).  

While impressive, this is a very constrained set of structures, nothing justifying the claims made in the popular press of solving all proteins. To be comprehensive, it seems that AI will have to consider biology, implement means of including the amino-terminal-first synthesis, nucleation, domain folding, insertion into a membrane, and above all interaction with chaperone proteins. 

Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Ε½Γ­dek A, Potapenko A, Bridgland A, Meyer C, Kohl SAA, Ballard AJ, Cowie A, Romera-Paredes B, Nikolov S, Jain R, Adler J, Back T, Petersen S, Reiman D, Clancy E, Zielinski M, Steinegger M, Pacholska M, Berghammer T, Bodenstein S, Silver D, Vinyals O, Senior AW, Kavukcuoglu K, Kohli P, Hassabis D. Highly accurate protein structure prediction with AlphaFold. Nature. 2021 Aug;596(7873):583-589. doi: 10.1038/s41586-021-03819-2. Epub 2021 Jul 15. PMID: 34265844; PMCID: PMC8371605.

Saturday, November 12, 2022

Leptin boosts protective vaccine responses

Follicular helper T cells (Tfh) support B cell development and production of antibodies, essential for a protective vaccination response. Metabolism has been linked to T cell development and the metabolic hormone leptin varies up to 10-fold among healthy people. Here, the investigators asked whether leptin levels might influence T cell development and contribute to variability in vaccine responses. 

Within a cohort of 76 healthy adults, they found non-responders to influenza vaccination had on average 2.5-fold lower serum leptin levels, with non-responders 10-fold more frequent in the low leptin group (fig 1ab). Tfh counts correlate with leptin levels (fig 2). Similar observations were made among older flu vaccine recipients (age >64 yr) and young Hepatitis B vaccine (HBV) recipients. Adding leptin to T cells cultured in vitro increased Tfh markers and production of IL-21 (fig 2e). 

In mice, they found leptin in areas of B cell development and leptin receptors on Tfh cells. Leptin receptor deficiency reduced antibody responses (fig 3b, c) and (consequently) allowed viral growth (panel a) in mice infected with H1N1 influenza. Tfh in leptin-receptor-deficient mice produced less IL-21 (fig 5b) and supplemental IL-21 restored most antibody production (fig 5a). IL-21 production is abrogated in T cells lacking STAT3 (fig 5g), strongly supporting a mechanism involving STAT3 and IL-21. 

Fig 7. Leptin protects from fasting-induced susceptibility to influenza. 


They could transiently reduce serum leptin levels by ‘fasting’ (starving) mice on alternate days 5 to 15 days after infection with influenza (Fig 7a, shown above). This timing chosen to avoid interfering with T cell priming (d 0-5) and focus on peak Tfh development (starting d5). Supplemental leptin protected against influenza (panel b), underscoring the significance of this pathway.

Deng J, Chen Q, Chen Z, Liang K, Gao X, Wang X, Makota FV, Ong HS, Wan Y, Luo K, Gong D, Yu X, Camuglia S, Zeng Q, Zhou T, Xue F, He J, Wei Y, Xiao F, Ma J, Hill DL, Pierson W, Nguyen THO, Zhou H, Wang Y, Shen W, Sun L, Li Z, Xia Q, Qian K, Ye L, Rockman S, Linterman MA, Kedzierska K, Shen N, Lu L, Yu D. The metabolic hormone leptin promotes the function of TFH cells and supports vaccine responses. Nat Commun. 2021 May 24;12(1):3073. doi: 10.1038/s41467-021-23220-x. PMID: 34031386; PMCID: PMC8144586.

Wednesday, November 2, 2022

Physical activity increases gut bacteria diversity

Previous work established associations in humans between physical activity and reduced obesity, reduced mortality, and improved cardiovascular health. Physical activity has been also associated with the microbiome in animals. Here, the relationship between physical activity and microbiome in humans was investigated.

The authors studied a cohort of 720 adults, citizens of Wisconsin, average age 55 years, 83% White, 10% Black, 42% male. Gut microbial, (bacterial) composition was assessed using sequencing the V3-V4 region of 16S rRNA extracted from stool samples. Note this is only a subset of the ‘microbiome’, not include non-bacterial components such as fungi, viruses, etc. 

They monitored physical activity using accelerometers worn on the hip (activity) or wrist (sleep). Participants also self-reported whether in a typical week they walked or biked at least 10 minutes continuously to get around. Those who responded ‘yes’ were classified as participating in ‘active transportation’. Note this is a threshold of less than half a mile a week, walking only about 100 m per day. 

Table 2. Linear mixed effects models (adjusted for characteristics, Table 1). CI, confidence interval; SD, standard deviation; MVPA, moderate to vigorous physical activity; ** p<0.05. *** p < 0.01.
They identified 865 unique bacterial taxa, largely encompassed by about 20 abundant phyla (Fig 1). They observed no change in bacterial diversity in participants who engaged in moderate-to-vigorous activity (line 2, Table 2, shown) or active transportation (line 3). However, when they analyzed those participants who engaged in higher levels of active transportation, at least 1 standard deviation (SD) above the average, they observed significant increases in bacterial diversity (line 4).

They also found the abundance of an unknown family from order Clostridiales was associated with increased weekly MVPA minutes. They conclude that their results “point to a potential pathway by which the gut micro- biota may be linked to physical activity and other well established health benefits”.

Holzhausen EA, Malecki KC, Sethi AK, Gangnon R, Cadmus-Bertram L, Deblois CL, Suen G, Safdar N, Peppard PE. Assessing the relationship between physical activity and the gut microbiome in a large, population-based sample of Wisconsin adults. PLoS One. 2022 Oct 26;17(10):e0276684. doi: 10.1371/journal.pone.0276684. PMID: 36288361; PMCID: PMC9605031.

Thursday, October 27, 2022

When viruses cohabit: Flu + RSV = Hybrid Frankenvirus

Experts worry that this winter might be made miserable by unwelcome visitors: something new, coronavirus variants, something flu (influenza A virus, IAV) and something blew, respiratory syncytial virus (RSV). What happens when somebody hosts IAV and RSV at the same time?

These investigators infected cultured human lung cells, A549 cells, and confirmed previous reports that coinfection reduces RSV but not IAV replication (Fig 1). Despite producing lower titer, they observed that infection with IAV appeared to increase the rate of coinfection by RSV.

Fig 3b. Filament with features of IAV and RSV.

IAV and RSV are enveloped viruses that bud from the cell membrane with characteristic glycoproteins hemagglutinin (HA) and fusion (F), respectively. Having detected HA in areas of RSV budding from coinfected cells, the authors hypothesized that some virions would contain components of both viruses. Indeed, they observed many filaments, typical of RSV, with proteins from both viruses, albeit segregated (Fig 2a-e). A remarkable scanning electron micrograph appears to show hybrid viral particles (HVP) budding from the filaments (2f, red arrows). They analyzed the hybrid buds using cryo-ET and were able to ‘segment’ features of both viruses (shown, Fig 3b): mostly IAV virions budding from mostly RSV filaments. 

Amazing biology, but what does it mean clinically? The authors found that the hybrid virions contained IAV capable of infecting cells that had been depleted of their sialic acids, which bind HA, by treatment with neuraminidase (NA), Fig. 4-5). This could be an important mechanism widening the range of infected cells.

Haney J, Vijayakrishnan S, Streetley J, Dee K, Goldfarb DM, Clarke M, Mullin M, Carter SD, Bhella D, Murcia PR. Coinfection by influenza A virus and respiratory syncytial virus produces hybrid virus particlesNat Microbiol. 2022 Oct 24. doi: 10.1038/s41564-022-01242-5. Epub ahead of print. PMID: 36280786.

Friday, October 21, 2022

Bad news bears on life choices (vaccine hesitancy)

Vaccine hesitancy, a reluctance or refusal to be vaccinated, probably began when Jenner invented vaccination over two centuries ago. Vaccines have largely eliminated scourges such as smallpox and polio and greatly reduced the rates of other infectious diseases including influenza. Anti-vaccination (anti-vax) stances stem from small, well-established risks of side-effects (managed by a compensation program) and big, vague worries about unrelated, even disproven associations with other maladies. COVID-19 vaccines were developed rapidly and rushed into production, potentially raising valid safely concerns.  However, any valid concerns were allayed when the COVID-19 vaccines were tested and proven safe and effective (Walsh 2020). 

The cable television show Fox News Channel (FNC) amplified concerns about COVID-19 vaccines and downplayed their benefits.  This study of viewership and vaccination covered ~2,750 counties (out of ~3,000 total) in 47 (of 50) US states documents that FNC viewers refused COVID-19 vaccination more often than the viewers of its competitors Microsoft-National Broadcasting Company (MSNBC) or Cable News Network (CNN) (Figure 2, shown).

Figure 2. Effect of network viewership on weekly vaccination rates, 2021

A key question is whether FNC influenced its viewers to refuse vaccination (a cause) or rather were anti-vax viewers attracted to FNC’s messaging, a consequence of playing to its audience.  The investigators used positions in cable channel listing as ‘exogenous shifters’ of viewership (Martin & Yurukoglu 2017).  Viewers are induced into watching more or less of a channel by variation in its position up or down the listing (Fig S3).  They found that “exogenously higher FNC viewership due to channel position causes lower vaccine uptake”.  They show that hesitancy was raised by FNC but not by competitors MSNBC or CNN (Fig 1). Moreover, resistance to vaccination against COVID-19 but not seasonal flu… causal…. Using the channels’ position in the guides.

Their “results imply that watching one additional hour of [FNC] per week for the average household reduces the number of vaccinations by 0.35–0.76 per 100 people”, which would account for a lot of ‘excess deaths’ in many households. Not surprising when “vaccine bad” was said so much more often on FNC than the other channels (Fig S7)! Although they found that FNC’s influence was mostly on those under 65 years old, who are at lower risk severe disease, those younger people are reservoirs of virus for infecting older people. Data-driven lawyers representing survivors of FNC victims could bring class action lawsuits. 

Pinna, M., Picard, L. & Goessmann, C. Cable news andCOVID-19 vaccine uptake. Sci Rep 12, 16804 (2022). 

Tuesday, October 18, 2022

Lipid metabolism and dementia

About half the human brain mass is lipid. Several brain disorders are known to be caused by abnormal lipid metabolism, disruptions in the processes of making lipids and breaking them down. Second only to Alzheimer’s in prevalence is frontotemporal dementia (FTD), one form of which encompasses a range of social, behavioral, or language disorders (as opposed to memory or motor deficits seen in other disorders).  Several genes have been associated with FTD, foremost among them the conserved genes MAPT (tau), PSEN1 (presenilin), VCP (valosin containing protein) and GRN (granulin).

Granulins are a family of secreted, glycosylated peptides (A, B, C, etc.) cleaved from a single precursor (progranulin, PGRN), that are involved in a wide range of activities probably due to their roles in regulating protein lysosomal protein metabolism.  The authors of this study found that gangliosides (i.e., glycosphingolipids with attached sialic acids) are elevated in brains of granulin mutant mice (GRN R504X), which are analogous to the most prevalent granulin mutation in humans, R493X (substitution of the arginine normally at position 493 with a nonsense codon, resulting in a truncated protein). This mutation causes neuronal ceroid lipofuscinosis, a severe neurodevelopmental disease, in humans and neuroinflammation in mice (Jax).  The metabolic order in the ganglioside degradation pathway (Fig. 1a) is first disialylated GD1 (Fig. 1b, shown, rightmost plot, annotated with red #1) πŸ‘‰ monosialylated GM1 (#2)  πŸ‘‰ GM2 (#3) πŸ‘‰ GM3. Also, GD2 πŸ‘‰ GM3 (#3’) via an alternative pathway. Levels of precursor GD1 (#1) are significantly elevated in the brains of mice with heterozygous mutant granulin (Figure 1b: blue fill, Grn +/R493X) compared with normal granulin (Grn+/+, grey) but not in homozygous mutants (purple). This may suggest a feed-back mechanism that limits the accumulation of that metabolite. GM1, #2, is elevated in homozygous mutant brain compared to normal, with the heterozygous mutant intermediate. However, GM2, #3, is not significantly elevated in mutation-bearing mouse brains. The alternate pathway, #3’, shows elevated GD3, the precursor to GM3, in homozygous mutant brains.

They also analyzed the lipids in postmortem human brains of 12 GRN mutation-related FTD cases, 6 sporadic FTD cases, and 3 control normal subjects.  GD3 and GM1 are significantly elevated in GRN-related FTD cases (Fig 1c, blue columns). However, they both seem also elevated in non-GRN (sporadic) FTD cases (green). GD1 is significantly elevated in GRN-related but not -unrelated FTD.  

In a striking simplification, they tested the effects of removing the granulin precursor protein gene, PGRN (same as GRN), in HeLa cells (Fig 2). They found elevated GM2 in the deficient line (GRN-/-) that was reduced to normal levels by restoring granulin (GRN-/- + PGRN-addback).

What causes changes in the levels of gangliosides? Gangliosides are catabolized by lysosomal enzymes. However, those enzymes were not altered by GRN deficiency (Fig 3). An intermediate metabolite, bis(monoacylglycero)phosphate (BMP), which is crucial to ganglioside degradation, was found to be reduced by 50-60% in GRN-deficient HeLa cells and mouse brains, and ‘markedly’ in human brains of FTD cases, although again both GRN-related and sporadic cases (Fig 4d). The authors propose a model wherein “lysosomal granulin peptides maintain lysosomal function and homeostasis, including the levels of BMP, that are crucial for ganglioside catabolism”. Their results await confirmation by others and many details remain to be pursued further. One relatively simple aspect will be clarifying how the model accounts for autosomal dominance of GRN deficiency. Also worth noting is the proximity of GRN and the Alzheimer- and Parkinson-associated MAPT genes, within a million nucleotides in band 17q21.31. 

Boland S, Swarup S, Ambaw YA, Malia PC, Richards RC, Fischer AW, Singh S, Aggarwal G, Spina S, Nana AL, Grinberg LT, Seeley WW, Surma MA, Klose C, Paulo JA, Nguyen AD, Harper JW, Walther TC, Farese RV Jr. Deficiency of the frontotemporal dementia gene GRN results in gangliosidosis. Nat Commun. 2022 Oct 7;13(1):5924. doi: 10.1038/s41467-022-33500-9. PMID: 36207292; PMCID: PMC9546883.

Thursday, October 6, 2022

COVID restrictions – Moderation is good

In efforts to reduce the spread of COVID-19, billions of people around the world were subject to rules and laws governing their behavior. Among the various non-pharmaceutical interventions (NPIs), what worked? 


An influential early report analyzing data from 11 European countries during the first 4 months of the pandemic found that lockdowns reduced transmission rates (Rt values) significantly, ~80%, whereas other NPIs such as cancelling public events, school closure, encouraging social distancing, and self-isolation, resulted in less significant reductions (0-20%) (Fig 2, Flaxman).  The dramatic drop in infections after lockdown is so obvious that it required no modeling (Fig 1).  A similar study found the most effective NPIs for lowering cases were travel restrictions, school closures, and the partial lockdown (Cortis).  A related study of 19 NPIs during seasonal flu found that banning large gatherings was most effective in limiting transmission (Qiu).  These studies used epidemiological models that directly involve underlying mechanisms.


Data on cases, deaths, vaccinations, and tests, were obtained from the COVID-19 Data Hub (Guidotti).  NPI data were obtained from the Oxford Covid-19 Government Response Tracker (OxCGRT).  


In this report, the author analyzed data from 132 countries between Feb 2020 to April 2021, capturing 3 waves of infection, beginning March 202, July 2020, and January 2021.  An econometric model with 4 equations: C = cases growth rate, D= deaths growth rate, M = mobility, and p(SI) = probability of the assigned stringency intensity level was employed.  Stringency correlates inversely with nonresidential mobility (Fig 3).  He found that ‘unobserved variables’ influence the growth of cases and deaths (C and D) as well as the stringency (SI) of government policies. Medium-stringency measures greatly reduced case and death growth rates but, surprisingly, yet-more-stringent measures slightly increased them (Fig. 4, shown).   

Fig. 4 Case (panel a) and Death (panel b) growth rates vs NPI Stringency Index. 

Testing helped but contact tracing did little (Fig 5).  The benefits of reduced nonresidential mobility were outweighed by increased within-household transmission.  Various differences in culture, compliance, and enforcement of government imposed NPIs were acknowledged but not clearly managed. Even low levels of vaccination reduced Case and Death growth rates with nonlinear improvement anticipated. 

The findings disagree with previous reports that lockdowns are effective, concluding that "very stringent NPIs provide no further benefits over moderately stringent ones, and that less stringent NPIs function primarily as signals for significant voluntary changes in citizens’ behavior.". Such analyses are crucial for designing effective, results-driven policies and for persuading people to comply.  


Spiliopoulos L. On the effectiveness of COVID-19 restrictions and lockdowns: Pan metron ariston. BMC Public Health. 2022 Oct 1;22(1):1842. doi: 10.1186/s12889-022-14177-7. PMID: 36183075; PMCID: PMC9526209.